Add some low-effort type annotations

[harahu]

This PR adds some uncontroversial annotations to the central trio.run method. From my perspective, it is mainly intended as an experiment into whether a gradual, opportunistic approach to adding types to this project is feasible.

[codecov]

Codecov Report

Merging #2477 (fc4ed29) into master (af3d7d8) will increase coverage by 1.04%.
The diff coverage is 94.11%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2477      +/-   ##
==========================================
+ Coverage   92.46%   93.50%   +1.04%     
==========================================
  Files         118      118              
  Lines       16327    16336       +9     
  Branches     3156     3157       +1     
==========================================
+ Hits        15096    15275     +179     
+ Misses       1103      955     -148     
+ Partials      128      106      -22     

Impacted Files	Coverage Δ
trio/_core/_run.py	98.92% <94.11%> (+0.22%)	⬆️
trio/tests/test_subprocess.py	96.96% <0.00%> (+0.55%)	⬆️
trio/_socket.py	95.68% <0.00%> (+0.78%)	⬆️
trio/_core/tests/test_io.py	99.29% <0.00%> (+1.05%)	⬆️
trio/tests/test_socket.py	98.21% <0.00%> (+1.13%)	⬆️
trio/tests/test_threads.py	100.00% <0.00%> (+1.56%)	⬆️
trio/_highlevel_socket.py	98.19% <0.00%> (+1.80%)	⬆️
trio/tests/test_highlevel_open_tcp_stream.py	100.00% <0.00%> (+1.83%)	⬆️
trio/_subprocess.py	93.91% <0.00%> (+2.17%)	⬆️
... and 8 more

[pquentin]

Thank you for reviving this effort! Previous attempts are https://github.com/python-trio/trio-typing, #1873 and #2208.

A gradual approach is certainly possible: https://sethmlarson.dev/blog/tests-arent-enough-case-study-after-adding-types-to-urllib3. But I think you have to make it work file by file using an allowlist, as done for urllib3. Case in point, the mypy check is failing for this pull request.

[harahu]

@pquentin We can probably figure out the exact mypy config as we go.

The reason it failed was due to the check_untyped_defs = False option being set. The moment I started adding types to a few previously untyped functions, I uncovered issues hidden in said functions. This is probably fine, though, as the alternative would be pretty noisy at the moment.

I've made mypy succeed now, but I ended up adding some uncovered lines in the process.

[A5rocks]

The reordering of imports is unfortunate, but this looks good overall.

Unfortunately compatibility with typing is a bit cursed: from __future__ import annotations is in limbo (note the lack of resolution for 3.12). If that ever gets removed, we need to make sure we use that version's types at runtime -- of course, it shouldn't be. It's likely that we'll have to change it to co_annotations or similar soon. We'll just have to take this as we can. A good next step would be adding mypy to ci.sh to run it on every platform we test (other than PyPy ones, of course).