bug(pipeline): digest never drains on channel close; events lost on shutdown #331

Problem

1. crates/trapfalld/src/digest.rs:47-67: Some(event) = self.rx.recv() only disables its select branch when the channel closes; the interval branch is always ready so else never fires and the loop never exits.
2. crates/trapfalld/src/main.rs:249-255: axum::serve has no with_graceful_shutdown; after it returns, digest_handle.abort() kills the task without a final flush.

Impact

Events already acked with HTTP 200 (up to 50 in batch + queued in the 1024 mpsc) are lost on SIGTERM/restart/deploy.

Fix

• match self.rx.recv().await { None => { flush; break } Some(e) => ... } inside the select.
• Handle ctrl_c/SIGTERM, graceful shutdown, drop ingest_tx, then digest_handle.await instead of abort().
• Add a test: send N events, close channel, assert all persisted.

Found in static audit of develop @ db3ca67.