POC: make ccalendar.pyx valid python

[jbrockmendel]

Proof of concept for making some of our cython files into valid python*, which would allow us to lint and maybe (cc @simonjayhawkins thoughts?) use mypy

* strictly speaking, we need to do a find-and-replace of "cimport" -> "import" to get all the way to valid python.

[simonjayhawkins]

cool. i've been looking into this but not got very far since I need to better understand cython first.

One thing that I have thought while investigating is that it may be possible to move some non performance related code (not looping) out of the cython.

for instance in array_with_unit_to_datetime in pandas/_libs/tslib.pyx, there seems to be logic to determine whether to iterate. I wonder whether code like this could be in python so that the cython code is just the iteration. (I think some more code that could be pure python being added in #35027?)

of course the approach in this PR helps us to keep relevant code together, while still allowing linting and mypy checking.

[WillAyd]

Is this style documented by Cython?

[jbrockmendel]

Is this style documented by Cython?

It's documented as "pure-python mode".

One thing that I have thought while investigating is that it may be possible to move some non performance related code (not looping) out of the cython.

As long as it doesn't break the self-contained-ness of tslibs and near-self-contained-ness of _libs, I'm all for this.

@simonjayhawkins is this something where we get a non-trivial benefit even if we only get a handful of modules? This is one of the easiest modules to do this on. In particular pure-python mode doesn't support const int64_t[:] or ndarray[int64_t] etc.

[jbrockmendel]

Closing to clear the queue since this is just a proof of concept. Can re-open if @simonjayhawkins thinks there is benefit to getting some but not all of these files python-valid