feat(worker): dispatch Qhorus COMMAND on worker channel after scheduling

[mdproctor]

Closes #186

What

When WorkerScheduleEventHandler successfully submits a worker execution, it now:

1. Opens a worker-specific channel: caseChannelProvider.openChannel(caseId, "worker:<name>")
2. Posts a COMMAND: caseChannelProvider.postToChannel(channel, "casehub-engine:orchestrator", json)

The COMMAND payload:

{
  "type": "COMMAND",
  "capability": "<capability-name>",
  "correlationId": "<event-log-id>",
  "input": { ... }
}

Why

Previously the engine provisioned workers and opened a case coordination channel but never sent a COMMAND. The entire Qhorus obligation lifecycle (ACKNOWLEDGED → FULFILLED/FAILED → trust score) was bypassed for all CaseHub-orchestrated work. There was no way to distinguish "agent acknowledged and is working" from "agent never picked up the work."

How cleanup works

Worker channels are tracked by the CaseChannelProvider alongside the case coordination channel. CaseStatusChangedHandler already calls listChannels() + closeChannel() for all open channels when a case reaches a terminal state — worker channels are cleaned up automatically with no additional wiring.

Test

Adds commandDispatchedToChannelWhenWorkerScheduled to SpiWiringIntegrationTest, which records postToChannel calls and asserts a COMMAND containing the capability name is sent.

Also fixes application-memory.properties missing H2 datasource config (ledger JPA entities need it even in in-memory test profile) and adds quarkus-jdbc-h2 to the persistence-memory Maven profile.

🤖 Generated with Claude Code

[treblereel]

/retest

[github-actions]

✅ Tests restarted — check the Actions tab