Multiscalar multiplication w/o precomputations

[survived]

We often have computations like:

$$Q = s_1 P_1 + \dots + s_n P_n$$

which can be computed more efficiently than calculating each $s_i P_i$ separately and $\sum$-ing them.

In this PR, I implement various techniques for computing multiscalar multiplication efficiently.

Note: docs fail to check in CI because some dependency fails to compile on nightly