This should at least make it clearer why/where stuff is failing.

Taken from
https://github.com/actions/cache/blob/main/tips-and-workarounds.md#force-deletion-of-caches-overriding-default-cache-eviction-policy.

Caches are per-branch, so if a PR is closed we do not need to keep any
caches for that branch around anymore. This PR adds automation to do
just that.

Enable sending events to spotlight by setting the `SENTRY_SPOTLIGHT` environment variable

---------

Co-authored-by: Burak Yigit Kaya <bkaya21@bloomberg.net>

…ted (#13278)

[ref](#12351)

This PR moves the `SentryGlobalFilter` out of the root module, which led
to the filter overriding user exception filters in certain scenarios.
Now there is two ways to setup sentry error monitoring:
- If users have no catch-all exception filter in their application they
add the `SentryGlobalFilter` as a provider in their main module.
- If users have a catch-all exception filter (annotated with `@Catch()`
they can use the newly introduced `SentryCaptureException()` decorator
to capture alle exceptions caught by this filter.

Testing:
Added a new sample application to test the second setup option and
expanded the test suite in general.

Side note:
- Also removed the `@sentry/microservices` dependency again, since it
does not come out of the box with every nest application so we cannot
rely on it.

This streamlines some caching stuff for CI:

1. Extract dependency installation & cache out into a composite action
for reusability
2. Updated the cache key for dependencies to only include package &
dev-package `package.json`, not the E2E test ones.

Instead of having to keep to separate lists of include/excludes, we now
keep this in a single list and invert it when necessary.

This way, we should no longer have problems where tests are either run
multiple times, or not in the correct env - just add the test to the
`browser` list in `ci-unit-tests.ts` to make sure it is not run in
multiple node versions unnecessarily. I also added this step to the new
package checklist.

🤞 

Closes: #13345

Should reduce amount of cached data.

This improves a few things in our size-limit CI action:

1. Show change in bytes, in addition to the change in percentage. 
2. Add a link below the table to the base comparison run.
3. If we detect that the workflow run we used as base is not the latest
one, show a warning on top.


![image](https://github.com/user-attachments/assets/4678ff04-a463-4579-ad91-74cbf9b7d781)

…13338)

This streamlines some stuff in our browser integration tests, to fix
some flakiness (hopefully).

The biggest change is that instead of always building into `dist` for
each test file, each test will now build into a random subfolder, e.g.
`dist/abc`. This way, multiple tests in a single file will never
conflict with each other.

Additionally it also streamlines some of the tests I encountered while
looking at stuff, hopefully reducing flakes further.

Closes #13321

Reverts #13118

There has been quite a bit of segmentation faults in CI after this PR
was merged.

I tried spending time to look into this in more detail, but it didn't
get far, so not worrying about this too much for now. Let's revert and
re-examine later.

relax the pageload transaction assertion. Web vitals are too
flaky to hard-assert on them and they are covered in browser integration
tests

Minor changes: I realized that the formatting was broken for `v8` and
that the migration guide wasn't linking to the specific section like
other changelogs.

CI is angery. Crypto is only on global from Node 19 onwards.

This PR adds the external contributor to the CHANGELOG.md file, so that
they are credited for their contribution. See #13395

Co-authored-by: mydea <2411343+mydea@users.noreply.github.com>