Add type annotations for @intrinsic

[aneeshdurg]

Previously using the intrinsic decorator would cause typing issues when invoking the decorated function, e.g.:

@intrinsic()
def myIntrinsic(typing_context, a, b):
   ...
   
myIntrinsic(a, b) # type checker complains that there's a missing parameter

This can be solved by annotating intrinsic with types that indicate that the decorated function does not take the typing context parameter.

[aneeshdurg]

the failure in the docs build seems to be because python version before 3.10 did not have typing.Concatenate. This can by resolved by either taking typing-extensions as a dependency, or possibly moving the annotations into a stub file and making it conditional on the python version. Will try to test that out soon.

[aneeshdurg]

I was able to resolve the CI issues by moving the annotations to a .pyi file. This allows importers of numba.core.extending to gain type information about intrinsic, but doesn't actually do any type enforcement within extending.py.

[aneeshdurg]

@stuartarchibald now that this is passing tests can the review process proceed?

[stuartarchibald]

@stuartarchibald now that this is passing tests can the review process proceed?

@aneeshdurg yes, I've now queued it for review. Thanks!