* Adding additional instrumentation to `publish-event!`

The `publish-events!` defmethod has quite a few code branches and we're trying to figure out if one of these is causing the slowness described in #33499 as well as stats dashboard [2117](https://stats.metabase.com/dashboard/2117). This PR should provide additional information as to where the delays might be occurring.

It adds the following spans/changes:
- An `:events/initialized` attribute at the top level of `publish-event!`
- `record-view!` span
- `record-event!` span
- `update-users-recent-views!` span

If we merge this PR we can decide if these changes are worthwhile and revert if we don't like the additional layer of wrapping.

* Reinstating record-event! span to resolve rebase conflict.

* Fixing bad merge

The `values-from-card-query` constructs a new query to determine
enumerable values for filter drop-downs in dashboards. It does this
by generating a simple query from the passed in query with the
filtered value as a query filter. This query didn't know about
expressions which the filter might use. The fix was to just add
in the expressions from the base query.

Fixes #12985

also adds quickcheck tests for user-id and model-id precedence

* Adding clojure goes fast dependencies and configuration for debugging slow queries.

* Adding OTEL Span Instrumentation

This PR adds a Clojure library for adding opentelemetry spans for
code instrumetation along with a couple examples of instrumented
areas that could be hot spots.

* Adding span around run-query-async

* Adding `publish-event-async!`

The evidence for the slow loads of dashboards such as https://stats.metabase.com/dashboard/2117
as described in https://github.com/metabase/metabase/issues/33499
point to long synchronous calls in `metabase.events/publish-event!`.

This change does the following:

- Adds a span to `publish-event!` for instrumentation
- Adds an async version of the function
- Changes the GET api/dashboard/:id call to use the async function

Based on the discussion in [this slack thread](https://metaboat.slack.com/archives/CKZEMT1MJ/p1697604232592589)
we may not want to retain the async call ATM, but we could use it as
a stopgap for now.

As an aside, we might want to introduce a queue such as activemq if
we really need asynchronous behavior.

* Rebasing and reconciling with master

* Modifying ws to be like master in deps.edn

* sort nses

* Removed dead code

* Removed unneeded check-404

* Added TODO comment wrt goals for db performance

* Whitespace changes

* Removing async publication call

Several parts of this structure contain MBQL and dashboard/card
references that need exporting and importing, which was being missed.

Resolves #33882

Prevent updating a database from closing connections unnecessarily when updating `database-enable-actions` setting (#35713)

* Improve and unify `external-filter` checks

* Refine E2E jobs conditionals

* Improve `uberjar.yml` workflow

It previously didn't even run multi-arch job if not on `master`.
There's no reason not to run this job. The trick is only to **not** push
the changes to DockerHub, unless we're on the `master` branch.

* Add fonts in Dockerfile to fix pulses (#31764)

Add Noto fonts in Alpine Dockerfile

* Fix Docker build command

---------

Co-authored-by: Luis Paolini <paoliniluis@gmail.com>

* add special flaky group

* tag tests with flakes over 10%

* mark more flaky jobs

This reverts commit 44de976d.

* Hide edit question menu item on dashboards when there is no collection write permission

* Add :can_write to dashboard->dashcards->cards

This hydrates each card in a dashboard's dashcards with :can_write.

* Check in a dashboard test that :can_write shows up on cards.

Just adding a key to one of these tests to let us know that it's being hydrated.

---------

Co-authored-by: Adam James <adam.vermeer2@gmail.com>

* fix: duplicated loadingIds on pending action

* chore: rename unit test name

Fixes #35719

* Adding linter and formatter to staged BE files.

Note that `.cljfmt/indents.clj` tries to capture our indent rules
and, so far, seems to get things right. There are still some inconsistencies
in some of our nses in how we indent assoc(-in) and some t2 functions
when it comes to key-value pairs on following lines. Should the keys
align with the function call or be indented. Aesthetically, people
seem to like indents, but these are functions, so should align with the
first argument. IDK that it really matters as long as we have agreement.

We may, as we adopt this, have some files be reformatted in unexpected ways
for items that have been missed. If so, the developer can simply update
the .cljfmt/indents.clj file, rerun the hook, and commit that along with
their other changes.

* Added whitespace linter.

* Updating husky commit hook

* Updated hook scripts to be exclusively for staged files.

Updated cljfmt to use the latest and greatest along with the new format of config file.

* Commenting out the actual formatting hook in .husky/pre-commit until we come up with a globally happy solution to forms that don't have good formatting rules, like defprotocol+ in a reader conditional.

* Reverting formatting

* Moving pre-commit hooks from .husky/pre-commit to the package.json's lint-staged section.

Note that we're still not calling the cljfmt file (.bin/cljfmt_staged.sh)
until we get agreement on formatting.

One thing we might do since we're using lint-staged is we can probably
exclude the "bad" files that go sideways with linting and format everything
else automatically. This might be a good follow-on PR.

Add Noto fonts in Alpine Dockerfile

* fix: misplaced negative sign

* update formatting unit test

* chore: update test name

* remove internal user on rollback + test

* fix naming + add newline

* fix typo

* rollback change to migration changeset ID

- since it would alter the checksum for it