design: SLA propagation — case budget should bound child WorkItem and Commitment deadlines

[mdproctor]

Problem

SLA deadlines exist independently at three levels with no relationship between them:

• casehub-engine: PropagationContext.deadline (case-level budget, propagates to sub-cases only)
• quarkus-work: WorkItem.expiresAt (set independently by the spawning caller)
• quarkus-qhorus: Commitment.expiresAt (set independently at COMMAND creation time)

A case with a 1-hour budget can legitimately spawn a WorkItem with expiresAt = +48h. The case will FAULT after 1 hour; the WorkItem will remain ASSIGNED for another 47 hours with no knowledge that its case has ended.

Similarly, a Qhorus COMMAND created during a 1-hour case can have expiresAt = +24h — the commitment will show as OPEN long after the case has faulted.

Proposed resolution

When the casehub-work-adapter (casehub-work-adapter module in casehub-engine) spawns a WorkItem, it should:

Instant effectiveDeadline = earliestOf(
    requestedDeadline,          // from CaseDefinition
    ctx.propagationContext().deadline()  // remaining case budget
);
workItemRequest.expiresAt(effectiveDeadline);

When ClaudonyCaseChannelProvider.postMessage() creates a Qhorus COMMAND (per engine issue #185), it should apply the same min(requested, case.budget.remaining) bound to Commitment.expiresAt.

The PropagationContext.createChild() already exists for sub-case budget propagation — the same concept needs to flow to WorkItems and Commitments at the integration layer.

Affected modules

• casehub-work-adapter (casehub-engine) — WorkItem spawn deadline bounding
• claudony-casehub — Qhorus COMMAND Commitment deadline bounding
• No changes needed in quarkus-work or quarkus-qhorus core

Part of

#4 (platform coherence audit — finding #18)