Add OHLCV caching to avoid yfinance rate limits in TA signal workflows

[Copilot]

The run_strategy tool fetches OHLCV data via yfinance on every execution when TA signals are used, causing slowness and potential rate limiting during iterative chat sessions.

Changes

• Parquet-based OHLCV cache under ~/.optopsy/cache/ohlcv/{SYMBOL}.parquet
• Smart range merging in run_strategy: fetches only missing date ranges, merges with cached data
• Cache CLI commands: optopsy-chat cache size now shows OHLCV stats, cache clear supports OHLCV category
• Gap detection: refetches interior gaps >5 calendar days (consistent with options chain provider logic)

Example

# First call: fetches 2024-01-01 to 2024-12-31 from yfinance
run_strategy("long_calls", "SPY", entry_signal="sma_crossover", ...)

# Second call with overlapping range: reuses cache, fetches only 2025-01-01 onward
run_strategy("long_calls", "SPY", entry_signal="rsi_oversold", start_date="2024-06-01", end_date="2025-03-01", ...)

Cache survives across chat sessions. Historical OHLCV data is immutable, so no TTL needed.

---

🔒 GitHub Advanced Security automatically protects Copilot coding agent pull requests. You can protect all pull requests by enabling Advanced Security for your repositories. Learn more about Advanced Security.