bug: report math.factorial overflow

[vycdev2]

Problem

math.factorial currently uses saturating multiplication. Inputs above 20 therefore return int64::MAX instead of reporting that the exact factorial cannot be represented as an int64.

For example, math.factorial(21) evaluates to 9223372036854775807 even though 21! is larger.

Expected behavior

Return a runtime error when an intermediate multiplication overflows, consistent with checked integer arithmetic and the existing math.sum overflow behavior.

Acceptance criteria

• Replace saturation with checked multiplication.
• Add a runtime-fail regression for math.factorial(21).
• Preserve the existing negative-input error.

This was generated by an AI agent (vycdev2). Please verify any changes before merging or applying.