Fix scheduler pending reason, dispatch requeue, and salloc timeout (#90, #91, #92)

[powderluv]

Summary

Fixes three reopened bugs: #90, #91, #92.

Issue #90: Jobs stuck PENDING with Reason=Priority

Root cause: Job::new() set initial pending_reason to Priority before the scheduler evaluated the job.

Before:

$ sbatch -N 1 test.sh
Submitted batch job 1
$ squeue
  JOBID  PARTITION  NAME  USER  ST  TIME  NODES  NODELIST(REASON)
      1    default  test   nod  PD  0:00      1  (Priority)      <-- WRONG

After:

$ sbatch -N 1 test.sh
Submitted batch job 1
$ squeue
  JOBID  PARTITION  NAME  USER  ST  TIME  NODES  NODELIST(REASON)
      1    default  test   nod  PD  0:00      1  (None)           <-- correct
# Next scheduler tick: job transitions to Running

Issue #91: Container jobs stuck in PENDING (Resources)

Root cause: When agent rejected dispatch (e.g., container image not found), job was immediately marked Failed. No retry.

Before: Container image not found -> job immediately Failed
After: Job requeued to Pending so user can fix the issue (import image) and the scheduler retries

Issue #92: salloc hangs with no interactive I/O

Root cause: salloc polled indefinitely for RUNNING state with no timeout or feedback.

Before: salloc hangs forever with no output
After: Shows pending reason, times out after 5 minutes with clear error

Test plan

• t07_50: Initial pending reason is None
• t07_51: Held jobs keep PendingReason::Held
• t07_52: Job schedules immediately on idle nodes
• t07_53: Container jobs schedule same as bare-process
• Full suite: 808 tests, 0 failures

@amd-kmundiga — please verify these fixes resolve the issues you reported. The key changes:

1. Pending reason now shows (None) initially instead of (Priority) — the scheduler sets the real reason after evaluating
2. Container dispatch failures requeue instead of failing immediately
3. salloc shows progress and times out instead of hanging

🤖 Generated with Claude Code