fix: job wrapper status now that the watchdog kills the payload

[aldbr]

Since #8416 the Watchdog kills the payload itself. A job stopped for exceeding a limit usually has no exit code, and postProcess reported it as "Application thread did not complete" / "No outputs generated from job execution", hiding the Watchdog's reason. executePayload then overwrote the minor status with "Exception During Execution", which is also what the monitoring and the accounting recorded.

• Without an exit code, postProcess reports the Watchdog's reason first, then an executor error, then the generic case.
• The CPU figures are propagated before any early return, so accounting sees them.
• executePayload, and the PushJobAgent after a remote postProcess, keep a Failed status that postProcess already set. A traceback is still logged when it did not.
• Unchanged on purpose: when the Watchdog killed the payload, its verdict stands even if the payload exits 0 or asks to be rescheduled on the way out. A new test pins this.

Case	Before (monitoring / accounting)	With this PR
Watchdog kills the payload, no exit code (time left, CPU, wall clock, disk, stalled, kill command)	Failed / Application thread did not complete is sent, then executePayload queues Failed / Exception During Execution, flushed by sendFailoverRequest. Final: Exception During Execution. The Watchdog's reason is only in the log.	Failed / (e.g. Job has reached the CPU limit of the queue). ApplicationError = the same reason, kept by executePayload, and in accounting.
Same, plus an executor error	Application thread failed, then the Watchdog reason, then Exception During Execution	The Watchdog reason, kept
Executor error, no exit code, no Watchdog	Application thread failed, then Exception During Execution. Message: No outputs generated…	Application thread failed, kept. Message: the executor's error
No exit code, no error	Application thread did not complete, then Exception During Execution	Application thread did not complete, kept
Watchdog killed it but the payload still returned an exit code (non-zero, 0, reschedule)	Failed /	Unchanged (0 and reschedule now pinned by a test)
Normal exits: 0, non-zero, reschedule	Application Finished Successfully / With Errors / rescheduled	Unchanged
postProcess never ran (e.g. Payload process could not start)	Exception During Execution + traceback	Unchanged
No CPU figures available	__sendFinalStdOut raised KeyError, ending as Exception During Execution	The final output heartbeat is skipped, and the status reflects the real outcome

Part of #8745.

BEGINRELEASENOTES
*WorkloadManagement
FIX: a payload killed by the Watchdog is reported with the Watchdog's reason (e.g. "Job has reached the CPU limit of the queue") instead of "Application thread did not complete" / "Exception During Execution", also in the accounting and for the PushJobAgent
ENDRELEASENOTES

[aldbr]

• Tested in certification

[fstagni]

Converted to draft as waiting for certification test.