LSP test-run projection still reads managed runs from the in-memory operation-event ring

[ScriptedAlchemy]

#2583 moved managed test runs to the durable managed_test_runs store (global-db), and tracedecay_test_results and Loom read that store. The LSP test-run projection still reads the same runs from the process-global operation-event ring (OperationEventAuthority, retained_event_capacity: 256):

• crates/tracedecay-application/src/lsp_runtime/managed_test_runs.rs OperationEventTestRunProjection → CanonicalManagedTestRunReader::latest_current_page → OperationEventAuthority::latest_managed_test_run (operation_stream.rs).
• After a daemon restart, or once 256 newer events evict it, the LSP answers FrontierExpired for a run that is durably recorded.

That leaves two readers for one fact. Migrating the LSP reader needs the durable row to carry what the LSP snapshot uses and the store lacks today: per-document document_content_digests (for staleness), plus the live completed/total progress while a run is still in flight. The ring may stay the live-progress stream, but the terminal result should come from managed_test_runs.

Found by the legacy-api-sweep lane; outside its area (LSP runtime).