[mypyc] Add str.isdigit() primitive

[VaggelisD]

Similar issue as the str.isalnum() PR, for large enough strings the primitive introduces a perf regression but is generally faster on common cases:

All-digit strings (100M calls each)	Python (s)	mypyc (s)	Speedup
length 1 ('0')	2.089	0.656	3.2x
length 10 ('1234567890')	2.475	1.028	2.4x
length 100 ('5' * 100)	6.106	3.406	1.8x
length 1 (UCS-2: U+0660 ٠)	2.110	0.734	2.9x
length 10 (UCS-2: U+0660 * 10)	2.907	2.041	1.4x
length 100 (UCS-2: U+0660 * 100)	10.596	13.887	0.8x

Non-digit strings (100M calls each)	Python (s)	mypyc (s)	Speedup
length 1 ('a')	2.022	0.585	3.5x
length 100 ('a' * 100)	2.068	0.587	3.5x
length 100 ('0' * 99 + 'a')	7.304	3.474	2.1x

[VaggelisD]

There's a recurring pattern for these primitives, should we try to abstract their codegen?

Gave macros a shot for to hide the per-kind for loop, though we could go a step further and do the same for entire functions I guess.

[JukkaL]

One option would be to have an inline function which gets passed a function pointer to represent the variable functionality, on the assumption that C compilers can simplify all the overhead away (not sure if this is the case, but it might well be).

Another idea would be to add a template for all of these functions in a comment at the top of the file, and we could just ask Claude Code or Codex to create another function based on the template for a new use case. And if we update the template, we could use a coding agent to update all instances of the template in the code. The problem with this is that there would be no automatic validation against things being consistent, but we could add some comments warning against manual edits.

[VaggelisD]

Not entirely sure why this specific CI/CD job fails here, seems to be a network issue?

Was probably one, fixed after rebasing.

[JukkaL]

Thanks for the PR! Looks good, but there are now some merge conflicts. Can you fix them?

[JukkaL]

One option would be to have an inline function which gets passed a function pointer to represent the variable functionality, on the assumption that C compilers can simplify all the overhead away (not sure if this is the case, but it might well be).

Another idea would be to add a template for all of these functions in a comment at the top of the file, and we could just ask Claude Code or Codex to create another function based on the template for a new use case. And if we update the template, we could use a coding agent to update all instances of the template in the code. The problem with this is that there would be no automatic validation against things being consistent, but we could add some comments warning against manual edits.