functools.lru_cache in the stdlib is not generic enough. That can be used to cache the results of a single pure function,
however I have usecases where one function is supposed to populate the cache, while another function consumes it.
https://docs.python.org/3/library/functools.html#functools.lru_cache
This is stripped down and extracts just the LRUCache from tkem/cachetools. It is relatively short, and very mature code.
I don't expect that we have to "follow" upstream, etc. There likely won't be relevant changes upstream.
Effectively, we are forking and bundling this code.
similar to 04582cc353
