Add Torch implementation (all classes excluding SL3)

[angadbajwa]

first pass at full torch implementation. making this now so you guys can start looking at it while I finish it up.

checklist of things left to add (that I know of):

• np cross-validation. maybe a quick addition to brute force a logm test in batch
• documentation. base class is definitely lacking relative to the np base.py and the dimensionality rules need to be explained so it doesn't get confusing
• edit np test_so3.py because I added ordering to the to_euler function
• dimensionality isn't fulllllly enforced. the rules are followed with inter/intra-function interactions, but we could idiot-proof this further. it's possible that the user forgetting that you need a batch dimension could lead to some funky behaviour especially in the lower-dimensional classes (i.e. SO2)

[CharlesCossette]

Just need to add torch to this list i think:

https://github.com/decargroup/pymlg/blob/main/setup.py#L83

[CharlesCossette]

Also, for the dimensionality thing, if its just something thats not supported right now you can throw a value error

raise ValueError("Input must be an [N x 2 x 2] matrix")

and write a comment that its a TODO or something