Stub Loading is Slow and Repetitive

[BradenM]

Currently, loading stubs requires validating the same stubs multiple times. Since the stub JSON file is read from on each validation, this is a very slow and repetitive process. When you have ~5 or more stub packages, executing micropy tends to have a very noticeable delay.

A much faster solution would be to only validate when a stub is added (not just loaded from the stubs folder) then create and save an md5 hash of the stub info file along with what type of stub it is. We can then "validate" and determine existing stubs by reading from this one file, rather than re-validating them all.