fix: prevent header times scan panic after stop

[lox]

Problem

headerTimesStreamer.Stop() closed timesCh but left streaming set to true.

Scan() checks streaming and, when true, sends a timestamp into timesCh. If Scan() runs after Stop() closes the channel, it can panic with:

send on closed channel

Relevant code:

• stop path: agent/header_times_streamer.go (Stop)
• send path: agent/header_times_streamer.go (Scan)

Impact

This is a runtime crash risk in job execution cleanup.

In practice, log-processing goroutines can still call Scan() while cleanup is in progress. If that races with Stop(), the agent process can panic instead of finishing job teardown cleanly.

Potential user-visible effects:

• abrupt agent failure during/after job completion
• incomplete cleanup and confusing job-end behaviour
• reduced confidence in log/header-time streaming reliability

Reproduction

Added a regression test that exercises the real lifecycle:

1. start streamer with Run()
2. call Stop()
3. call Scan("--- a header")

Before this fix, the test fails with panic (send on closed channel).

Fix

Set streaming = false under streamingMu before closing timesCh.

This preserves the existing send/close synchronisation while ensuring post-stop Scan() calls do not attempt channel sends.

Testing

• Added: TestHeaderTimesStreamerScanAfterStopDoesNotPanic
• Verified with:
  • go test ./agent -run TestHeaderTimesStreamerScanAfterStopDoesNotPanic -count=1