Speed up augmentations

[svteichman]

Try to speed up estimation under alternative, especially for large n

Do this by:

• faster implementation of X_cup_from_X
• faster implementation of get_G_for_augmentations
• option to use mclapply for get_augmentations for settings in which plenty of memory is available and n is large, slowing down computation of augmentations

[svteichman]

From testing these updates, I think that X_cup_from_X_fast and get_G_for_augmentations_fast are speedups from their original functions. However, using get_augmentations_par, which parallelizes augmentations over n, even for very large n (in my example case n = 1880 does not give a speedup. I would suggest not using this (right now, the default is to not run in parallel). However, I will update this PR to remove this option, while keeping the fast helper functions.

[svteichman]

Closing this PR because it is out of date, but these fast X_cup and G generation algorithms are implemented in PR #180