fix(taskctl): heartbeatActiveAgents not transitioning developing→reviewing after binary restart

[randomm]

Problem

After a binary restart and taskctl resume, Pulse restarts and ticks every 5 seconds. heartbeatActiveAgents checks isSessionActivelyRunning(task.assignee) for tasks with stage: developing. The developer session is dead (not in memory after restart), so isSessionActivelyRunning should return false, triggering the developing → reviewing transition.

However, the transition is NOT happening — tasks remain stuck at stage: developing indefinitely.

Suspected Causes

1. Instance context mismatch: startPulse captures instanceContext.tryGet() at call time. After a restart, the captured context may not match the live instance, causing SessionStatus.get() inside isSessionActivelyRunning to throw and return false... but something else prevents the update.

2. pulse_pid display bug: taskctl status shows Pulse PID: none because it reads job.pulse_pid from the DB, but the PID is stored in a lock file. These are never in sync. Low priority cosmetic issue but causes confusion.

Acceptance Criteria

• After binary restart + resume, developing tasks with dead sessions advance to reviewing within one Pulse tick (5s)
• Adversarial agent spawns after transition
• taskctl status correctly reflects Pulse running state
• bun test and bun run typecheck pass

[randomm]

Resolved by PR #255 (merged ec304a5). heartbeatActiveAgents now uses isPidAlive(assignee_pid) as authoritative liveness signal alongside SessionStatus.