GH-93179: Document the thread safety of functools.lru_cache

[rhettinger]

• Issue: Document the thread safety of functools.lru_cache #93179

[miss-islington]

Thanks @rhettinger for the PR 🌮🎉.. I'm working now to backport this PR to: 3.11.
🐍🍒⛏🤖

[miss-islington]

Sorry, @rhettinger, I could not cleanly backport this to 3.11 due to a conflict.
Please backport using cherry_picker on command line.
cherry_picker eea8f42deddb3b6fec80984e7557b1f0c4578b62 3.11

[1st]

Sorry, but we have an issue with the @lru_cache as it doesn't work like a singleton like we expected.

In the code it has thread lock, but problem that it's done outside of the function call.

with lock:
# ... result = user_function(*args, **kwds)

As result, we have issue with threads - function call made multiple times, but we expected that this @lru_cache decorator will ensure that we receive exactly cached value that will be calculated only once.

Currently I'm in a doubt, should we create our own version of the decorator that will have the lock over the whole @lru_cache. Maybe you did it by purpose, by not having the function call inside the lock?