Caching helper functions broadly do not work

[russellp17]

Describe the bug
...
#210 added some caching helper functions including set_cache_size, get_cache_size, and clear_cache but these functions don't really work as intended.

dacite/cache.py defines a cache decorator and then the decorator itself is cached using @lru_cache(maxsize=None) - this layer of "meta" caching does not have any effect. @cache is invoked as a decorator on many functions, but only once - the Python interpreter applies the decorator once per decorated function as it parses the library. You can see the effect using cache_info():

>>> dacite.cache.cache.cache_info()
CacheInfo(hits=0, misses=17, maxsize=None, currsize=17)

These 17 misses and currsize correspond to the 17 functions decorated - 15 in dacite/types.py and 2 in dacite/dataclasses.py. But there are no cache hits and it would be difficult to even make use of these (I think you'd have to manually call cache(f.__wrapped__) to reapply the caching to the original, un-decorated function).

clear_cache clears the above "meta" cache, but this has no effect on the underlying cached functions like dacite.types.is_generic - in order to clear those caches you'd have to go through them like dacite.types.is_generic.clear_cache() - you could in theory have the cache function maintain a global list of cached functions so that clear_cache could iterate through this and clear them.

set_cache_size sets a global variable but this has no effect on the caching behavior - dacite.types.is_generic would have already been decorated by the time any user is calling set_cache_size and you can see by inspecting it with dacite.types.is_generic.cache_info() that the maxsize does not get updated. My understanding is that there is no good way to update a cache size dynamically.

To Reproduce
...

Expected behavior
...

Environment

• Python version: ...
• dacite version: ...

Additional context
Add any other context about the problem here.