test: skip stale Redis status entries

[chico10117]

Summary

• make bridge-operation status scans skip Redis set members whose operation record has already disappeared
• return a clear error for unknown status names instead of scanning an empty Redis key
• add miniredis-backed regression tests for stale status-set members and unknown statuses

Validation

• go test ./redis
• go test ./...
• git diff --check

Supports the testing/reliability goals in #3.

Preferred payout if accepted: USDC on Base / EVM wallet 0x820a7bf90d944bb26bfD9b62Ab172Fc3A0829cB9.

[MyTH-zyxeon]

Review assist for maintainers:

This looks like a focused Redis consistency hardening patch: stale status-set members whose backing operation key has expired or been deleted now get skipped instead of crashing the lookup path, and unknown status names return a clear error before SSCAN. The new miniredis tests cover both FindAllBridgeOperationsByStatus and source-tx lookup behavior with stale set members.

Suggested merge checks:

• Run go test ./redis or the repo's normal Go test target and confirm miniredis remains a test-only dependency.
• Verify both lookup paths still surface real Redis errors after the new redis.ErrNil skip branch, so connection/protocol failures are not hidden as stale records.
• Consider whether stale set members should also be cleaned from the Redis set after skip; if not, the current behavior is still a safe read-path fix and can be merged separately from cleanup semantics.

I do not see wallet/key, bridge signing, live RPC, or live-chain behavior changes in this patch as written; the main acceptance criterion is that stale Redis membership no longer breaks bridge operation lookup while valid records still deserialize and match normally.