MontgomeryPoint Basepoint multiplication Table and performance

[elichai]

EdwardsPoints have EdwardsBasepointTable, I have a usecase where I use the whole Montgomery curve (wheras with the Edwards I'd need to also implement the twist curve) by utilizing #555, And I actually have 2 constant basepoints one on the curve and one on the twist that I'm using for multiplying with random scalars.
I'd like to create some kind of statically precomputed table for those 2 points,
I've also recently stumbled upon this paper/note: https://eprint.iacr.org/2017/264 which allows to further improve the precomputation of MontgomeryPoint Tables.

There's also things like https://eprint.iacr.org/2020/388 to improve the regular Montgomery Point multiplication (See #588)