fix(hub): pass resumeSessionId when resuming session

[xyzhang626]

Summary

Fixes a resume argument wiring bug in SyncEngine.resumeSession() that prevents hub-triggered session resume from passing the actual resume token to the CLI spawn RPC.

Root cause

rpcGateway.spawnSession() takes params in this order:

1. machineId
2. directory
3. agent
4. model
5. modelReasoningEffort
6. yolo
7. sessionType
8. worktreeName
9. resumeSessionId

In hub/src/sync/syncEngine.ts, resumeToken was passed as the 8th argument (worktreeName) instead of the 9th argument (resumeSessionId).

So for resume flow:

• hub found metadata.claudeSessionId (or other flavor token)
• but CLI never received resumeSessionId
• runner spawn command missed --resume <id>
• session could start as a new session instead of resuming the target one.

Fix

In hub/src/sync/syncEngine.ts:

• add explicit undefined for worktreeName
• pass resumeToken as the 9th arg (resumeSessionId)

Tests

Added regression test in hub/src/sync/sessionModel.test.ts:

• passes resume session ID to rpc gateway when resuming claude session
• verifies engine.resumeSession() forwards metadata.claudeSessionId into rpc spawnSession(..., resumeSessionId)

Also keeps existing model-forwarding resume test intact.

Validation

• cd hub && bun run test src/sync/sessionModel.test.ts
• all tests in this file pass

Impact

• fixes resume correctness for Claude (and same call path shared by other flavors)
• prevents silent new-session fallback caused by missing resume token at spawn RPC boundary

[github-actions]

Review mode: initial

No issues found in added/modified lines. Residual risk: resume flow not exercised end-to-end beyond unit test coverage. Testing: Not run (automation).