fix: bind to correct port for e2e tests

[hiltontj]

• The TestServer now binds to port 0 so that the host will assign a random port, then scrapes the logs to get what port was assigned. This will prevent flakes in our E2E tests in CI that have been occurring from time to time.
• Changes initial log message on serve command start for naming from OSS to Core
• Includes the bind address along with the server startup time log message

Closes #25423

[hiltontj]

Just force pushed a small change to the catalog log output to match https://github.com/influxdata/influxdb_pro/pull/353

[hiltontj]

Ah, sadly it looks like CI does not work with this change. Tests are all hanging: https://app.circleci.com/pipelines/github/influxdata/influxdb/43145/workflows/24f834e5-ca0b-4520-9957-e9d13bc78107/jobs/405026

Will need to re-evaluate.

[hiltontj]

Okay, looking good again after 26a82f7

[pauldix]

Odd, why does that fix it?

[hiltontj]

Odd, why does that fix it?

It was because the TEST_LOG env var's value was being used as a log filter to the spawned server process, and passing anything that prevents the INFO log containing the server address breaks the change made by this PR (it hangs waiting to see the log with the server address in it).

I think it would make sense to make TEST_LOG not do that for this PR, since we need the logs to emit INFO from the influxdb3_server crate to capture the assigned port and that is a bit of a foot gun.

[hiltontj]

I opened #25760 - I don't think fixing that is a high priority - I felt like getting this in to prevent CI flakes from blocking is more important.