Make the node database operate on jobs instead of pods

[zuqq]

It secretly always operated on jobs, because PodRequirementsFromLegacySchedulerJobs was adding annotations containing job IDs and queue names to pods. This PR makes this behavior explicit: methods like ScheduleMany now operate on JobSchedulingContext (which holds the cached PodRequirements value) instead of plain PodRequirements.

[codecov]

Codecov Report

Patch coverage: 70.21% and no project coverage change.

Comparison is base (d6aef28) 58.78% compared to head (0d82d1f) 58.79%.

Additional details and impacted files

@@           Coverage Diff            @@
##           master    #2612    +/-   ##
========================================
  Coverage   58.78%   58.79%            
========================================
  Files         238      238            
  Lines       30541    30397   -144     
========================================
- Hits        17954    17872    -82     
+ Misses      11219    11177    -42     
+ Partials     1368     1348    -20     

Flag	Coverage Δ
armada-server	58.79% <70.21%> (+<0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
internal/armada/server/lease.go	7.53% <0.00%> (+0.02%)	⬆️
internal/scheduler/common.go	85.71% <ø> (+0.21%)	⬆️
...ernal/scheduler/schedulerobjects/schedulinginfo.go	0.00% <0.00%> (ø)
internal/scheduler/context/context.go	33.00% <26.08%> (-0.51%)	⬇️
internal/scheduler/jobdb/job.go	78.63% <33.33%> (+0.22%)	⬆️
internal/scheduler/pool_assigner.go	72.41% <76.19%> (-3.05%)	⬇️
internal/scheduler/nodedb/nodedb.go	61.15% <78.87%> (+2.73%)	⬆️
internal/scheduler/scheduler.go	78.12% <87.50%> (+0.33%)	⬆️
internal/scheduler/submitcheck.go	73.22% <91.66%> (-1.78%)	⬇️
internal/scheduler/gang_scheduler.go	82.55% <100.00%> (+4.43%)	⬆️
... and 4 more

... and 1 file with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.