Monte Carlo Dropout fastai v2

I will probably make a PR for this, it is very useful.

Good idea, I just made one: https://github.com/fastai/fastai/pull/2733

I noticed the current implementation of the MCDropoutCallback has before_validate and after_validate implemented. Is MCDropoutCallback available working during inferece of test data once training is complete?

This is more of a question trying to understand the callbacks before_validate and after_validate and if it is applied during test time.