runtime/lava: add tree-based priority support

[bhcopeland]

Add 4-level priority system for LAVA job queue ordering:

• highest (90): manual submissions for bisection/debugging
• high (75): critical trees (mainline, stable, stable-rc)
• medium (50): important trees (next, stable-rt, kselftest)
• low (25): other trees (default)

Human submissions (submitter != SERVICE_PIPELINE) get highest priority by default, but can optionally specify priority via node['data']['priority'] using string ('high'/'medium'/'low') or numeric values.

Pipeline submissions use tree_priority from build configs.

Add SERVICE_PIPELINE class constant to replace hardcoded string.

Priorities are scaled to each lab's configured range before submission to accommodate different LAVA lab configurations.

[broonie]

The commit log mentions the allocation of priorities to the trees but the changes don't actually do that, they look up the priorities from the config. That's good and sensible but it'd probably be good to have the actual config update here as an end to end example of things working and to help with testing on staging.

[bhcopeland]

The commit log mentions the allocation of priorities to the trees but the changes don't actually do that, they look up the priorities from the config. That's good and sensible but it'd probably be good to have the actual config update here as an end to end example of things working and to help with testing on staging.

Thanks, Broonie, for the feedback. The patch is also linked to another PR: kernelci/kernelci-pipeline#1407, should have linked it in this PR. This includes the config changes. I just triggered a staging run to see how it behaves. Let me know if I am missing anything or if priorities need adjusting.

[broonie]

Ah, didn't see that other PR - I checked for already merged stuff but didn't think to check pending PRs.

[nuclearcat]

As we discussed in Discord - this PR are well tested, so merging right now.