Trial ledger and suggested fixes

Trial ledger: --ledger

monte-neo verify --ohlcv data.csv --strategy s.py --ledger

Each verification appends a row to .monte-neo/ledger.jsonl (an append-only file, every row hashes the previous one). A variant is identified by its normalised source (docstrings stripped) or by the hash of its positions, scoped by the hash of the data. The number of distinct variants tried on this data replaces what you declare when it is larger: n_trials = max(declared, counted). Changing a parameter until the result looks good therefore raises the Deflated Sharpe bar by itself. Pass a path to keep the ledger elsewhere. MCP: verify_strategy(ledger=true).

Suggested fixes: --suggest-fix

monte-neo verify --suggest-fix strategy.py

Rewrites the patterns that read the future into causal ones and prints a diff; nothing is written. Handled: negative shift, center=True, bfill, interpolate, whole-sample statistics (to expanding), whole-sample rank, group transform('max'/'min'/'sum'/'prod'). Chains that use tail, head, iloc, values and similar are left alone. The patch keeps the shape of the strategy, not its meaning, and comments are lost: run verify on the result. MCP tool: suggest_fix.