Fix integer overflow check in multiply_frac

[Amit-Matth]

Description:
This PR replaces the fragile 32-bit overflow detection logic in multiply_frac (in resample.c) with safer, standard 64-bit arithmetic.

Changes:

• Implementation of multiply_frac updated to cast operands to unsigned long long before multiplication.
• Verified overflow detection by comparing the result against UINT32_MAX.
• Removes the complex and potentially incomplete conditional logic previously used to guess overflows.

Motivation:
The original code contained a TODO comment acknowledging the method was suboptimal: /* TODO: Could use 64 bits operation to check for overflow. But only guaranteed in C99+ */. Since modern build environments support C99/64-bit integers, this change improves code safety and maintainability, ensuring that buffer sizes for the resampler are calculated correctly without wrapping.

Related Issues: #88

[joseph-yiu]

Thanks for this pull request.
Looks fine to me. I will ask my colleague to have a look as well.
"unsigned long long" could be more than 64-bit in some architecture but that shouldn't be an issue.
regards,
Joseph

[joseph-yiu]

Note: my colleague highlighted that resampling is not used in AudioMark. But the fix looks fine.

[joseph-yiu]

Merged into dev branch (dev_2026q1).