fix(foundation): add execution event parity for workflow lifecycle

[Bhanudahiyaa]

Summary

Align workflow runtime event emission with ExecutionEvent schema for retry/failure and pause/resume lifecycle paths.

Motivation

Observability/debugging currently misses key execution transitions, causing a mismatch between declared event contracts and emitted runtime data.

Changes

• Added lifecycle event variants in ExecutionEvent:
  • WorkflowPaused
  • WorkflowResumed
• Updated WorkflowExecutor to emit:
  • WorkflowPaused on wait/HITL pause
  • WorkflowResumed on resume_with_human_input and resume_from_checkpoint
  • NodeRetrying based on retry count
  • NodeFailed on terminal node failures
• Corrected workflow terminal event logic:
  • WorkflowCompleted now emitted only when status is Completed (not Paused)
• Added regression tests:
  • retry + eventual success emits retry + completed
  • terminal failure emits NodeFailed
  • pause/resume emits WorkflowPaused + WorkflowResumed

Related Issues

#1562

Testing

• cargo test -p mofa-foundation workflow::execution_event -- --nocapture
• cargo test -p mofa-foundation workflow::executor::tests:: -- --nocapture
• cargo test -p mofa-foundation test_emits_ -- --nocapture
• cargo check -p mofa-foundation
• cargo clippy -p mofa-foundation --all-features -- -D errors
• rustfmt --edition 2024 --check crates/mofa-foundation/src/workflow/executor.rs crates/mofa-foundation/src/workflow/execution_event.rs

Checklist

• cargo fmt --check passes (for touched files)
• cargo clippy --workspace --all-features -- -D errors equivalent attempted (mofa-foundation target; repo currently has known -D errors lint naming issue)
• cargo test coverage for changed paths
• Architecture layer rules respected
• Small, focused PR scope

[Bhanudahiyaa]

@yangrudan @lijingrs

This PR focuses on event contract parity at the workflow runtime boundary.

ExecutionEvent already modeled retry/failure semantics, but executor emission didn’t consistently include those transitions. I kept the change intentionally narrow to mofa-foundation workflow event schema + executor emission points, and added regression tests to lock behavior for:

• retry + success
• terminal failure
• pause/resume lifecycle

[yabhimanyu2007-sys]

I reviewed the tests in this PR and they cover the main lifecycle paths well.

I’d like to contribute by adding a few additional edge case tests:

• Event ordering validation (e.g., retry events occur before NodeFailed)
• Ensuring no duplicate NodeFailed emissions
• Behavior when max_retries = 0 (no retry case)

These should help strengthen correctness guarantees for execution event sequencing.

Please let me know if this direction works.

[Bhanudahiyaa]

@yabhimanyu2007-sys
Thanks for taking a look, glad the coverage makes sense!

Those are good edge cases, especially around ordering and duplicate emissions. For this PR, I tried to keep the scope focused on establishing baseline lifecycle parity (retry/failure + pause/resume) and locking in the core transitions.

I think the cases you mentioned would be great as a follow-up focused on event sequencing guarantees (e.g., ordering + dedup invariants), since that’s slightly orthogonal to just emitting the missing events.

If you’re interested, feel free to open a separate PR building on top of this, happy to review 👍