In-place cloning of Montgomery form

[xuganyu96]

Originally meant to allow boxed Monty to be cloned without additional allocation. Also includes a change to the Monty trait to expose clone_from_montgomery to generics

[tarcieri]

Why not just use Clone::clone_from?

[xuganyu96]

Why not just use Clone::clone_from?

My bad for not realizing that the option is here.

[tarcieri]

Note that the implementation of clone_from isn't quite in-place yet the way this PR was.

We'd need to deal with mismatched numbers of limbs to make it work like what you had (in which case panicking is probably fine, so clone_from doesn't have weird performance edge cases)

[xuganyu96]

Note that the implementation of clone_from isn't quite in-place yet the way this PR was.

We'd need to deal with mismatched numbers of limbs to make it work like what you had (in which case panicking is probably fine, so clone_from doesn't have weird performance edge cases)

I originally raised this PR because of entropyxyz/crypto-primes#41, in which fjarri mentioned the Lucas testing having a lot of .clone() calls. After running a cargo flamegraph, however, I found that the bulk of the time of generating prime is spent on instantiating MontyForm so maybe allocating is not the bottleneck there.