Add durable saga & compensation handling in scheduler

[yaront1111]

Summary

Implement durable saga/compensation handling in the scheduler, aligned with CAP v2.0.16 job status updates.

Scope

• Add saga manager to record compensations and execute rollback on FAILED_FATAL.
• Update scheduler and workflow engine handling for FAILED_FATAL/FAILED_RETRYABLE.
• Update docs/wiki for CAP + saga semantics.
• Add unit tests for saga manager and related control-plane behavior.

Notes

• CAP dependency pinned to v2.0.16.
• Core coverage target (80%) not yet met; currently ~66% core coverage.