nala: Rework PRIMITIVE_TYPES check

[thebolt]

The old check wasn't appropriate as the order of qualifiers isn't enforced by pycparser nor C compilers, yet Nala enforced the order.

This is a problem with stdatomic.h which contains instances of "short unsigned int", a type not recognized by Nala.

Instead just check that all qualifiers and the core types are common, leaving the qualifier validation to pycarser and the compiler. Utilize a set for this.

Encapsulate the PRIMITIVE_TYPES list in a method and call that instead of implementing the set check everywhere.
This results in some changes to is_primitive_type_pointer and is_primitive_type_pointer_type to no longer take a types argument, instead they call the new is_primitive_type method as necessary.

[eerimoq]

Seems some tests fail. Not sure if it's due to the changes in the PR though. Do you? =)

python -m unittest passes on my computer, but fails in CI. hmmm

[thebolt]

It does seem very weird and somehow more related to the CI environment than the actual code.

I tested to run locally with python 3.9 (3.9.21, latest provided by uv and almost same as CI), 3.10.12, 3.11.11 and 3.14.0a4, and all of them passes.

I'm not really familiar with the github actions environment so cannot really say what the problem might be.