I'm not sure if this is something you want but it was a simple change so
I figured I might as well contribute it. This PR allows the user to
upload `.blob` files that they may have gotten from server's
`user-files/` folder. This can be useful if the user didn't export the
file but has server backups.

---------

Co-authored-by: Jed Fox <git@jedfox.com>

Disabling the flaky unit test step. We should re-enable it eventually,
but right now it just creates unnecessary noise..

Depends on server change:
https://github.com/actualbudget/actual-server/pull/168

Making consistent height between multi/single input autocomplete.

Enabling the new autocomplete for dev/preview deployments.

This will allow us to spot any more issues there might be before we
release the new autocomplete.

https://github.com/actualbudget/actual/issues/773

The final `Autocomplete` refactors. After this is merged what's
remaining is to do extensive testing and address the bugs in
https://github.com/actualbudget/actual/issues/773

This PR moves `Nordigen` autocomplete to the new one without using a
feature flag. IMO this is a safe change given the simple nature of the
Nordigen autocomplete component.

Refactored all feature flags to use the new `useFeatureFlag` hook.

Also added a new functionality to this feature flag: ability to define
custom "default" value for a feature flag. This will allow us to enable
the new autocomplete component for everyone using Netlify builds
eventually (need to address some issues before doing so).

Further iterations on the new autocomplete.

1. Created `AccountAutocomplete`
2. Started using new autocomplete in `GenericInput` (used for notes
field)
3. Extracted common functionality between the three new autocompletes to
a generic component: `Autocomplete`

- Add a “hide decimal places” setting to visually hide the `.xx` from
currency values globally
- When hiding the fractional digits, slightly decrease character spacing
to allow more digits to show up

Ref: #327

New settings:
<img width="566" alt="Screenshot_2023-03-17 14 19 46"
src="https://user-images.githubusercontent.com/25517624/225986815-b884b93d-02f9-48b3-a73d-d27f90b678cf.png">


Before/after:
<img width="149" alt="Screenshot_2023-02-27 21 47 07"
src="https://user-images.githubusercontent.com/25517624/222916856-21ab4f03-56c6-4b24-8fc1-ac4b883138b7.png"><img
width="131" alt="Screenshot_2023-02-27 22 02 01"
src="https://user-images.githubusercontent.com/25517624/222916859-cf882ca3-6087-4994-818e-239c3374e412.png">

I noticed that the first run flow is suboptimal for people who want to
import an existing file from Actual/YNAB. I’ve moved the welcome modal
into the management app and set it up to appear when there are no
budgets available (which also has the benefit of allowing people to see
the modal again!)

I think there’s some weirdness around getting the modal to reappear when
deleting a budget file which I want to work out before merging this.

This PR also reorganizes the management app a bit to reduce usage of
modals (currently, hitting escape while the budget list is open leaves
you with a blank page).

<img width="539" alt="Screenshot_2023-03-18 08 53 54"
src="https://user-images.githubusercontent.com/25517624/226107462-b2b88791-1015-4397-b290-c64e7fcc0f41.png">

- [x] Ensure modal consistently appears when needed (no longer a modal!)
- [x] Fix e2e tests

Added an extra `waitFor` after a flaky unit test step.

I'm not really super happy with this workaround.. but it does make the
test much more stable (re-ran 5x and no failures:
https://github.com/actualbudget/actual/actions/runs/4455134799).

I think there is some internal timeout happening somewhere which is
causing this issue.. But not really sure where. And this will hopefully
get auto-fixed once we migrate to a new table. 

Just cleaning up things: removing an unused code-path.

Brining back `onHover`. This is a small regression.

Finishing off the React v18 upgrade by doing a change in `react-dom`.
Effectively this upgrades from v17 to v18.

https://react.dev/blog/2022/03/08/react-18-upgrade-guide

Upgraded `fast-check` to improve unit test performance.

This is not a full fix for the flakiness. One of the test cases will
still be flaky. But at least this fixes the other test cases thus
improving stability.

I am currently not working on adding support for importing to multiple
accounts, but I wanted to give anyone who takes that on a starting point
by updating the underlying C library to provide access to the account
name field.

Tests cases should be independent. You should be able to run them in
whatever order you want. And they should still pass.

Currently this is not the case. The order of the tests is very important
due to the "pseudo" randomization algorithm.

This PR makes the mock data IDs truly unique thus better exposing the
issue in our tests. Also this PR fixes the dependency issues thus making
each test case truly independent.

---------

Co-authored-by: Jed Fox <git@jedfox.com>

This way formatting issues won’t prevent the preview from building (and
it should build a bit faster due to not having to run ESLint).

Removing
- Debugger
- `perf-deeets`
- `codemirror`

Thanks @kidglove57 for spotting this issue and helping me track down the
cause!

---------

Co-authored-by: Matiss Janis Aboltins <matiss@mja.lv>

See https://github.com/actualbudget/docs/pull/129

 for more details. If
this is accepted, I’ll fill in release notes for the PRs that have been
submitted since the last release and submit a corresponding PR to
`actual-server`.

---------

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

before:

```
kcab.worker.4bdc73a8d45eb2115156.js (2.1 MiB)
xfo.kcab.worker.4bdc73a8d45eb2115156.js (1010 KiB)
```

after:

```
kcab.worker.39f5fba82d7bc7477962.js (1.41 MiB)
xfo.kcab.worker.39f5fba82d7bc7477962.js (1000 KiB)
```

What’s changed:

- `loot-core` did not have a `browserslist` config, so
`@babel/preset-env` assumes we want to [transpile all the way back to
ES5](https://babeljs.io/docs/options#no-targets). I’ve removed the
`browserslist` config from each of the `package.json` files and moved it
to the root so this doesn’t happen again.
- I updated the target from `electron 3.0` to `electron 12.0` to match
our Electron dependency
- I’ve added `defaults` (currently equivalent to `> 0.5%, last 2
versions, Firefox ESR, not dead`) which is [recommended by
browserslist](https://browsersl.ist/#q=defaults). We could consider
tightening this, but it doesn’t offer a ton of space savings at this
point to just target Electron 12.
- Since much less transpilation will be happening, stack traces (dev and
prod) will be much easier to read!

Fixes
https://github.com/actualbudget/actual/issues/724#issuecomment-1468453526

And add the missing fallback condition that wasn't catered in
https://github.com/actualbudget/actual/pull/745

Why not `23.3.1`? Because we already released that version for
`actual-server`

Related to:
https://github.com/actualbudget/actual/issues/724#issuecomment-1455160250

Depends on https://github.com/actualbudget/docs/pull/126 to be merged
first.

Two changes here:
1. show "link account" only if actual-server is used (user is "online");
2. allow linking accounts only if Nordigen is configured (using new API
to get the status for it);

Also ported the `CreateAccount` modal to a functional component.

Fallback to `bookingDate` if `valueDate` is empty.

Fix for:
https://github.com/actualbudget/actual/issues/724#issuecomment-1464914546

Route aggregate queries in transaction grouped mode through the correct layer to remove deleted transactions (#247)

I recently migrated my personal usage of Actual over to the open-source
version and imported a bunch of transactions. I have a _lot_ of history
in Actual, including a lot of weird edge cases like deleted split
transactions. While reconciling I noticed that my account balance shown
at the top was incorrect, even though the running balance was current.

Digging into this, I discovered that we aren't correctly handling
aggregate queries when querying transactions in the "grouped" mode.
Aggregate queries don't make sense in the "grouped" mode. Grouped means
that you want a list of transactions that include both the parent and
child transactions (when they are split). If you are summing up all the
amount, you only want to consider non-parent transactions. So we switch
it back to "inline" mode, but the way we did this previously was to
manually stitch the query together.

Even though was add SQL to ignore deleted transactions, we still
possibly include them. A child transaction may not be marked as deleted,
even though the parent transaction is deleted. When a parent transaction
is deleted, all child transactions should be considered deleted as well,
regardless of their tombstone status. This is what the
`v_transactions_internal_alive` view does. Previously we weren't going
through this view though, so we could still potentially include split
transactions even though they've been deleted.

This is little hacky, but it fixes the immediate problem. We fall back
to the inline mode by modifying the where clause, and we also adjust the
view that it queries to use the correct one.