Preserve integer precision in ordering comparisons #17

Ordering comparisons have the same precision problem as equality: converting large integers to float64 can make distinct values indistinguishable.

{{ 9007199254740992 < 9007199254740993 }}
{{ 9007199254740993 > 9007199254740992.0 }}

Both expressions should return:

True
True

The integer values are within the supported range. The precision is being lost during comparison.

I’ll open a PR shortly that keeps integer–integer comparisons in integer arithmetic and handles mixed integer–float comparisons separately. The mixed path accounts for floating-point bounds and fractional values without converting the integer operand to a rounded float.

I'll raise PR and include regression tests for both examples, and I’ll link it back to this issue.