Fix divide-by-zero in aromaticity index calculation

[robertyoung3]

Summary

• _calc_ai() and _calc_ai_mod() in MolecularFormulaCalc divide ai_n / ai_d without guarding against ai_d == 0
• This raises ZeroDivisionError for molecular formulas where C - 0.5*O - N - S (or C - O - N - S for modified AI) equals zero
• Returns 0 when denominator is zero, consistent with the existing clamping to [0, 1]

Before

ai = ai_n / ai_d

After

ai = ai_n / ai_d if ai_d != 0 else 0

Applied to both _calc_ai() (line 563) and _calc_ai_mod() (line 604).

References

• Koch and Dittmar, 2006: https://doi.org/10.1002/rcm.2386
• Correction: https://doi.org/10.1002/rcm.7433

Test plan

• Verified both methods have the same pattern and same fix
• A denominator of zero means no aromatic carbon skeleton is possible, so AI = 0 is the correct semantic value

[kheal]

Good find. According to the literature, if either the ai_n <= 0 OR the ai_d <= 0, then ai (or ai_mod) should be 0. Can you adjust to encapsulate these scenarios more generally? That will fix the the divide-by-zero bug as well.

[robertyoung3]

Thanks for the guidance! Updated to check both ai_n <= 0 and ai_d <= 0 — if either is non-positive, AI is set to 0. This follows Koch and Dittmar (2006) more faithfully and covers the divide-by-zero case as a subset.

The if ai < 0 clamping is now redundant (when both ai_n and ai_d are positive, AI is always non-negative), so I removed it to keep the logic clean. The if ai > 1 cap is retained since the ratio can still exceed 1.

Applied identically to both _calc_aromaticity_index() and _calc_aromaticity_index_mod().