Fix: re-propagate interrupt causes in sendTurnForThread catchCause handler

[cursor]

Problem

The inner Effect.catchCause wrapper around sendTurnForThread in ProviderCommandReactor.ts unconditionally caught all causes, including interrupts. During shutdown or fiber cancellation, this would:

1. Create a spurious provider.turn.start.failed activity
2. Attempt to dispatch a command to the possibly-shutting-down orchestration engine
3. Swallow the interrupt signal before the outer processDomainEventSafely handler could re-propagate it

Fix

Added a Cause.hasInterruptsOnly(cause) check in the inner handler (matching the pattern used in processDomainEventSafely and elsewhere in the codebase) to re-fail with the cause when it is interrupt-only, allowing proper interrupt propagation.

  

Note

Re-propagate interrupt causes in sendTurnForThread catchCause handler

In ProviderCommandReactor.ts, the catchCause handler for provider turn start now checks Cause.hasInterruptsOnly(cause) and re-fails with Effect.failCause(cause) instead of recording a provider.turn.start.failed activity. Non-interrupt failures continue to produce the failure activity as before.

^{Macroscope summarized ae48a85.}