fix(indexing): recover_index_queue raises "'int' object has no attribute 'lower'" on every backend boot #9784

Observed

Every backend boot logs:

initialization.lifespan - WARNING - Index queue recovery failed (non-fatal): 'int' object has no attribute 'lower'

The exception is swallowed at autobot-backend/initialization/lifespan.py:1252 (_recover_index_queue), so queued indexing jobs are never recovered after a restart (the #1717 feature is silently broken).

Root cause (to trace)

Entry point: api/codebase_analytics/scanner.py:195 recover_index_queue(). An int is being .lower()-ed — almost certainly a value read back from Redis (where everything is bytes/str) is already an int, or a field expected to be a string is an int. Likely a deserialization/type-coercion mismatch when restoring the queued job payload.

Fix

Trace the .lower() call inside recover_index_queue (and helpers), coerce/validate the offending field, and add a regression test that round-trips a queued job through Redis and recovers it. Do not leave the broad except Exception → WARNING masking it.

Impact

Index-queue recovery after restart is non-functional. Discovered while triaging compose-startup log spam (#9771).