fix: handle divide-by-zero in Spark mod/pmod with ANSI mode support

[davidlghellin]

Which issue does this PR close?

• NA.

Rationale for this change

Spark's mod and pmod functions return NULL on integer division by zero in legacy mode (ANSI off), but DataFusion's implementation always threw a DivideByZero error regardless of the ANSI mode setting.

What changes are included in this PR?

Add ANSI mode support to spark_mod and spark_pmod via enable_ansi_mode config option
In legacy mode (ANSI off): division by zero returns NULL per-element
In ANSI mode (ANSI on): division by zero throws an error (unchanged behavior)
Add try_rem helper that handles per-element zero-divisor masking for integer arrays

Are these changes tested?

Yes:

• 18 unit tests in modulus.rs (including new tests for both ANSI modes)
• Updated pmod.slt and mod.slt sqllogictests with ANSI on/off coverage

Are there any user-facing changes?

Yes — mod(10, 0) and pmod(10, 0) now return NULL instead of erroring when enable_ansi_mode = false (the default), matching Spark behavior.