Implement MSF paper

[IgorSusmelj]

Mean Shift for Self-Supervised learning introduces a simple extension to BYOL.

Code is available here: https://github.com/UMBCvision/MSF

From paper:

[philippmwirth]

Looks like all we need is the adapted loss as the architecture is the same as for MoCo.

[pranavsinghps1]

I would like to hop on it!