Fix retention hanging forever when the permanent store is disabled

[damilolaedwards]

Summary

purgeOldBeaconBlocks queues every aged out block with the permanent store and then waits on a channel that the permanent store only closes once the block is actually handed off for processing. When the permanent store is disabled, which is the default, or stopped, or its queue is full, QueueBlock returned without ever touching that channel, so the wait never completed.

Since all six retention purges run one after another on a single goroutine, this stalls cleanup for every data type the moment the first beacon block ages out, silently and permanently. No attacker or unusual configuration is needed, just the documented default.

QueueBlock now closes the channel on every path, not only the one where a block is actually queued. Retention's wait also now selects on context cancellation as a second line of defense, so a purge in progress during shutdown cannot block indefinitely either.

Changes

• pkg/server/service/indexer/permanent_store.go: QueueBlock closes ProcessedChan on every early return path via a small shared helper, also used by processBlock's existing close
• pkg/server/service/indexer/retention.go: the wait on ProcessedChan now also selects on ctx.Done()
• pkg/server/service/indexer/retention_test.go: two new tests, no prior test coverage existed for retention at all

Test plan

• go build ./... and go vet ./... pass
• New tests pass with the fix applied
• Confirmed the tests actually catch the bug: reverted the fix, reran them, watched both fail with a timeout, then restored the fix and confirmed they pass again
• Pre-existing Docker-dependent tests elsewhere in this package could not be run in this environment (no Docker available), unrelated to this change