Rebalance native CI jobs so that Misc4 is not twice the duration of other jobs

[holly-cummins]

This PR aims to help these problems:

• The Misc4 job takes much longer than the other native jobs, meaning feedback on that job is slower
• The slow feedback matters because the Gradle project is in the job, and it seems to be one of the more likely projects to fail
• The Misc grouping name doesn’t give much insight into what’s in the bucket
• The Native Devtools job is so short it's a bit of a waste of a machine

To address these, I moved the gradle tests to the Devtools job, since it’s one of the fastest. Then I noticed that almost everything that was left was observability-related, so I renamed the job and distributed the non-observability projects through the other jobs. I also moved maven in with gradle, since they seem like a logical grouping.

I think there are a few risks with this kind of change

• Syntax errors causing build failures, because the PR will not exercise the new yaml. To mitigate, I tested on my fork: https://github.com/holly-cummins/quarkus/actions/runs/12884686079
• Jobs getting accidentally removed. I think we can only mitigate this by careful inspection (and maybe by keeping rebalances small and incremental so they’re easier to inspect).
• Name changes causing confusion for people looking at jobs, or for develocity analysis. I think this is not too big a problem.
• Making the balance worse. To check this, I made a change in core, which I’m hoping would clear the caches, and ran my test build.

@brunobat could you validate the contents of the observability bucket look like a complete-ish consistent set?

Here’s the timings I observed on my test runs. The codebases are a bit different between the two, but it shows the ratios.

Old

Misc 1: 31m
Misc 2: 22m
Misc 3: 30m
Misc 4: 54m (often it’s over an hour for me)
DevTools: 4m (!)

New

Misc 1: 26m
Misc 2: 29m
Misc 3: 28m
Observability: 35m
Build tools and DevTools: 29m

[brunobat]

Thanks for this @holly-cummins!

[geoand]

@brunobat you might want to rebase #43831 once this is in

[holly-cummins]

io.quarkus.observability.test.LgtmResourcesTest.testTracing is failing in both native and JVM jobs, so I'm pretty confident it's not related to these changes.

[geoand]

Yeah, that's been pretty flaky

[holly-cummins]

Looks like it started failing on January 17, in fact.

[brunobat]

Will take a look at that test

[holly-cummins]

Hi @gsmet, are you able to look at this?

[gsmet]

@holly-cummins it's on my list for today. Wanted to get @brunobat's PR first as one of them was conflicting and this one is the easy one.

[gsmet]

(I'm currently having a look at the LGTM timeouts)

[holly-cummins]

I'll do the rebase against #43831.

[gsmet]

Makes a lot of sense, thanks!

[gsmet]

I merged as the native tests passed.