v0.4.1 — Comprehensive Accuracy Audit

What's Changed

Fixes 13 data accuracy issues discovered during a comprehensive codebase audit:

Financial Accuracy

• Net Debt formula now subtracts short-term and long-term investments (not just cash)
• Current Ratio returns NaN when current liabilities = 0 (previously returned 0.0, which falsely implied zero liquidity)
• ROE returns NaN for zero equity (consistent with Debt-to-Equity behavior)
• ROE/ROA alerts now fire for negative values, not just low-positive
• Margin percentages (Gross, Net, Operating) return NaN when revenue = 0

Forecasting

• Forecast output validated for NaN/Inf — falls back to naive (last known value) instead of propagating bad data
• Fallback uses last known value instead of misleading 0.0 when models fail or data is insufficient
• Warning suppression narrowed from blanket ignore to specific convergence/UserWarning categories

Data Integrity

• Synonym partial-match sort key made consistent between find_synonym_value and find_best_synonym_row
• pd.isna() replaces np.isnan() to prevent TypeError on None in capex computation
• Quarterly FCF calculation skips periods with missing operating CF instead of silently using 0.0
• CSV export preserves full numeric precision (K/M/B formatting only in console)
• CLI rejects directories for --csv path

Tests

• 7 new test cases covering zero-denominator ratios, NaN/Inf forecasts, negative ROE/ROA alerts, investment-adjusted Net Debt
• 3 existing tests updated for corrected NaN behavior

Full Changelog: v0.3.0...v0.4.1