Add a 30-minute in-memory cache keyed by the complete Financial Modeling Prep upstream URL so repeated profile, metrics, earnings, and screener requests do not burn the free-tier quota. The cache keeps non-OK responses out of storage so transient rate-limit or provider errors can recover on the next request.
Refs: docs/AUDIT_REDESIGN.md item C2.
Co-Authored-By: GPT-5 Codex <noreply@openai.com>
