gh-3108: Avoid materializing f_locals for KI protection

[graingert]

Fixes #3108
Fixes #2670

In this pull request, we avoid materializing f_locals by using weak references to code objects instead.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.62%. Comparing base (21ea257) to head (6e97f9a).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3110      +/-   ##
==========================================
+ Coverage   99.58%   99.62%   +0.03%     
==========================================
  Files         121      122       +1     
  Lines       18166    18340     +174     
  Branches     3268     3281      +13     
==========================================
+ Hits        18091    18271     +180     
+ Misses         52       47       -5     
+ Partials       23       22       -1     

Files with missing lines	Coverage Δ
src/trio/_core/_ki.py	100.00% <100.00%> (ø)
src/trio/_core/_run.py	99.02% <100.00%> (-0.01%)	⬇️
src/trio/_core/_run_context.py	100.00% <100.00%> (ø)
src/trio/_core/_tests/test_ki.py	99.77% <100.00%> (+1.93%)	⬆️
src/trio/_core/_tests/test_run.py	100.00% <100.00%> (ø)
src/trio/_tools/gen_exports.py	95.90% <100.00%> (ø)

[A5rocks]

Plenty of nitpicks or me being confused; please don't assume I know what I'm doing with these comments. If the question doesn't make sense, it's cause I don't understand and not some weird encoded question.

[graingert]

It looks like this will fix #1752

and maybe in conjunction with #3112 it will fix this #2454

[A5rocks]

A bunch of really minor comments. This could merge as is without any changes IMO.

[A5rocks]

Nevermind, I was thinking about whether a deprecation is possible but it isn't because the problem is functions having ki protection only sometimes.

[graingert]

pre-commit.ci autofix

[A5rocks]

Just to be explicit, since all my comments were fixed and all my questions answered.