[Gitflow] Merge master into develop

This PR adds a safeguard to ensure we do not flush (=send) a replay that
is either too short or too long.

We allow to configure a `minReplayDuration`, which defaults to 5s and
maxes out at 15s. Whenever we try to flush and the duration is shorter
than this, we'll just skip flushing.

Additionally, we also skip flushing if the replay is longer than
MAX_SESSION_LIFE + 5s (=60min + 5s). This _should not_ happen,
technically, but apparently it still does. So while we figure out the
root cause of this, we can at least avoid sending stuff in that case.

Due to the [way pnpm handles
versions](pnpm/pnpm#6463), the version `*` we
use in E2E tests is actually incorrect, because it installs the _lowest_
version it can find, e.g. 0.1.0. This is usually not a problem as when
we use verdaccio, there is only a single version in the repository.
however, when running things locally/debugging stuff, and you run `pnpm
install` without verdaccio, stuff fails. This updates the versions used
in E2E tests to `latest || *`, which results in a correct resolution.

Playwright's event listeners and `page.goto` functions can occasionally
end up in race condition, even when they are invoked in the correct
order.

The workaround is to invoke them in `Promise.all`, unless there's a
specific need to separate them.

Aggregate exceptions' `exception.value` strings
weren't truncated correctly in our event preparation pipeline. The
tricky part here is that the `LinkedErrors` integration adds the
linked/child exceptions to the event in an event processor which is
applied only after we truncate the original event. Hence, we need to
also truncate the values in the event processor itself.

This was causing drift between profiling and transaction timelines

meta(changelog): Update changelog for 7.60.0