Move RavenDb id generation out of the common persistence seam

[rbev]

Summary

• Stop assigning RavenDB-shaped IDs in the shared Audit ProcessedMessage constructor. The RavenDB ingestion unit of work now assigns the existing ProcessedMessages-{processing-started-ticks}-{processing-id} document ID before storing the message or its body.
• Let InMemory assign a storage-neutral processing ID for its existing deduplication check. No new field is added to ProcessedMessage.
• Preserve existing RavenDB document IDs, attachment keys, and /messages/{id}/body URLs; add a RavenDB test for the ID format and body lookup.

Related Error-instance PRs

• Move all ravendb specific ID generation out of the shared project #5649 — moved RavenDB-specific ID generation out of shared Error persistence models and into the RavenDB persister; direct precedent for this Audit change.
• Derive failed error import id from the message, not a random Guid #5652 — introduced deterministic failed-error-import keys and deduplication coverage; related neutral-identity precedent, not a change made by this PR.