Use assembly calls for AES, optimize PRG

[tholop]

Try to optimize the PRG with a lightweight assembly call to AESNI instead of using the aesni crate:

sycret/src/le.rs

Line 142 in 368276c

// TODO: optimize format, byte operations, assembly call to AESNI (e.g. https://github.com/gendx/haraka-rs/blob/master/src/intrinsics.rs)

Here is an example of a Rust crate doing that (for a slightly older edition of Rust): https://github.com/gendx/haraka-rs/blob/e487b2821bfd5485e1e3df95909bdade2f615f3a/src/haraka256.rs#L6

Any other optimization of the MMO PRG is welcome! For instance, depending on the expansion factor needed, we might be able to save a few calls to AES:

sycret/src/le.rs

Line 153 in 368276c

// TODO: A 3x expansion PRG is slightly overkill

[tholop]

For reference, here is some interesting work: Lightweight Techniques for Private Heavy Hitters

• PRG evaluation optimization: see Appendix C.2
• Lightweight Rust code adapted from the aesni crate: https://github.com/henrycg/heavyhitters/tree/main/src