Fix negative-volatility guard silently skipped in FX option greeks - #276

Bug

FXVanillaOption.gamma, .vega, and .theta in financepy/products/fx/fx_vanilla_option.py each guard against a negative volatility with:

if np.any(volatility) < 0.0:
    raise FinError("Volatility should not be negative.")

This checks the truth value of np.any(volatility) (a Python bool) against 0.0, not each element of volatility against 0.0. Since a bool is never < 0.0, the condition is always False, so the guard can never raise. A negative volatility silently flows through into the pricing math and produces a wrong (non-error) greek instead of raising, while the sibling method delta() (which correctly uses np.any(v < 0.0)) does raise for the same input — an inconsistency within the same class.

Root cause

Operator precedence: np.any(x) < 0.0 reduces first, then compares the reduced boolean; the intended check is np.any(x < 0.0), which compares element-wise first, then reduces.

The identical pattern existed in financepy/utils/helpers.py:input_time, in the ndarray branch's check that a curve date isn't before the curve's value date (np.any(t) < 0), with the same effect: negative time-in-the-past inputs on an array of dates silently pass instead of raising.

Fix

• fx_vanilla_option.py: changed all three occurrences to np.any(volatility < 0.0) / np.any(vol < 0.0).
• utils/helpers.py: changed np.any(t) < 0 to np.any(t < 0).

Verification

• Added test_negative_volatility_raises in tests/unit/test_FinFXVanillaOption.py, asserting gamma/vega/theta raise FinError for a negative-volatility BlackScholes model.
• Red/green: reverted the fix and confirmed the new test fails against the old code; reapplied the fix and confirmed it passes.
• Ran the full unit test suite (pytest tests) before and after: 1050 passing before (plus the new test failing), 1051 passing after — no regressions, and existing positive-volatility results for gamma/vega/theta are numerically unchanged (spot-checked against test_vega_theta's Hull example values).