GH-134: Add Validity functions

[zeroshade]

Adding is_null, is_not_null, and is_nan functions and tests to the Compute implementations for Go.

This also removes the -ffast-math option from the SIMD optimized compiler Makefile so that our comparisons properly handle NaN values (-ffast-math assumes that no values are NaN and removes the checks). is_nan can be easily implemented for float and double values by dispatching to A != A.

Tests are also added for the new functions.

Ultimately the driving force behind this is as these are necessary for implementing scanning reads in iceberg-go

[kou]

Could you check CI failures?

[kou]

isNullKns or isNullOrNotNullKns?

[zeroshade]

renamed kernels.IsNullKernels() to kernels.IsNullNotNullKernels() to make it more obvious

[kou]

Ah, sorry. I mentioned a local variable name not a function name here.

[zeroshade]

@kou it took a while, but i figured out the cause of the failures. it was the upgrade to the new version of xstrings. v1.4.0 of xstrings.ToCamelCase capitalizes the first letter of the word, while v1.5.0 of ToCamelCase doesn't capitalize the first letter. They really should have marked that as a breaking change :( I think i'll look for something other than xstrings to perform the camel casing there so we don't have to rely on an old version of a package

[zeroshade]

@kou I don't know why the pre-commit is failing though, it passes just fine when I run locally on this branch.

Also for some reason it's not showing the output here so i can know what the actual failure is caused by, which is weird. When i run locally, if there's failures on the linting it outputs it properly to the terminal.

[kou]

The lint failure was caused by mixing Go 1.22 and 1.23. See also: golangci/golangci-lint#4908

I've pushed a commit to ensure using Go 1.23 for pre-commit.