The cache client is now the KV client (core/cache/kv.py, kv_client),
configured by KV_URL. rate_limit_check allows the request and logs when KV
is unreachable, and the RPC cache treats an outage as a miss. cloud_billing
reads KV directly instead of a missing usage_tracker.redis attribute; the
NFT refresh enqueues through an arq pool.
Co-authored-by: Hanzo Dev <[email protected]>
