ASYNC113 & ASYNC121 false alarm with nested sync functions

[jakkdl]

Fixes #286

I also did a search for visit_FunctionDef and visit_AsyncFunctionDef in the codebase and caught the same error for async113 (though it should very rarely occur for 113). At first glance a couple other visitors had the same problem, until I noticed those were cst visitors (which instead has an asynchronous attribute).

For 113 we could consider having the logic for visit_FunctionDef being the same as visit_AsyncFunctionDef instead of always setting self.aenter = False in case users are transforming sync methods to async methods, but given such exotic stuff we're maybe better off staying away and avoiding super obscure false positives.

We could consider adding a test that checks if any AST visitors only has one of the visitors defined, and could do the same for With/AsyncWith and For/AsyncFor (the latter which was in fact only caught in code review for #282). It will have a decent false-positive ratio though, but not very troublesome to have an ignore list or declare dummy methods.

Another approach is preferring cst over AST, and I've been inconsistent when creating new visitors despite #124. I should try and compile pros/cons and write a comment over there.

[Zac-HD]

Thanks @jakkdl!