Fix MolecularFormula._calc_kmd ignoring kendrick_rounding_method

[robertyoung3]

Summary

• MolecularFormulaCalc._calc_kmd() uses int() to derive the nominal Kendrick mass, which always truncates toward zero (equivalent to floor for positive values)
• This ignores the user-configurable kendrick_rounding_method setting (floor/ceil/round) that MSPeakCalc._calc_kmd() already respects
• Now reads kendrick_rounding_method from the parent peak's settings (or MSParameters.ms_peak as fallback) and applies the same floor/ceil/rint logic

Before

nominal_km = int(kendrick_mass)  # always truncates (floor)

After

if kendrick_rounding_method == "ceil":
    nominal_km = ceil(kendrick_mass)
elif kendrick_rounding_method == "round":
    nominal_km = rint(kendrick_mass)
elif kendrick_rounding_method == "floor":
    nominal_km = floor(kendrick_mass)

Context

MSPeakCalc._calc_kmd() (for experimental m/z) already respects this setting. MolecularFormulaCalc._calc_kmd() (for theoretical m/z) was inconsistent. With the default kendrick_rounding_method="floor", the behavior is unchanged since int() and floor() are equivalent for positive values.

Test plan

• Default behavior (floor) is unchanged — int() and floor() produce identical results for positive Kendrick masses
• Settings access follows the same pattern as kendrick_base (parent peak settings → MSParameters fallback)
• Includes hasattr guard for safety when _mspeak_parent is not yet set