[neat_periodic_task] Add heartbeat locks to NeatPeriodicTaskScheduler

[sigurdm]

Summary

This PR adds heartbeat support to NeatPeriodicTaskScheduler in package:neat_periodic_task.

Motivation

Currently, when a periodic task is claimed by a worker instance, the status file is set to state: 'running' with started: now. If that worker is terminated abruptly (e.g. SIGKILL, container replacement, App Engine / GCE VM shutdown, OOM):

• The status file remains in state: 'running'.
• Other workers observe state: 'running' and assume the task is still executing. They will back off and not retry until _timeout has elapsed.
• For tasks configured with long timeouts (e.g. 12 hours) to accommodate worst-case execution times, an early crash prevents any other worker from reclaiming the task for up to 12 hours, potentially triggering monitoring alerts.

Changes

1. NeatTaskStatus:
   • Added optional DateTime? heartbeat property.
   • Fully backward-compatible JSON serialization and deserialization.
2. NeatPeriodicTaskScheduler:
   • Added optional heartbeatInterval and heartbeatTimeout constructor parameters.
   • Defaults: If heartbeatInterval is provided, heartbeatTimeout defaults to 3 * heartbeatInterval. If only heartbeatTimeout is provided, heartbeatInterval defaults to heartbeatTimeout ~/ 3.
   • While _task() is executing, a periodic timer updates the heartbeat timestamp in _statusProvider.
   • In _iteration(), if status.state == 'running', other schedulers check whether the heartbeat has expired (now.difference(status.heartbeat) >= _heartbeatTimeout). If expired, the worker is presumed dead and the task is reclaimed immediately without waiting for the full _timeout.
   • trigger() also reclaims the task if the heartbeat is expired.
3. Tests:
   • Added tests verifying heartbeat updates during execution, reclamation of abandoned tasks after heartbeat expiration without waiting for _timeout, and trigger reclamation.