Refactor the Merkle Test into Crates

[iamrecursion]

Currently the Merkle test is a multi-file test that demonstrates many functionalities of the library. As part of M3 we want to refactor it to consist of multiple crates instead to prove out the dependency management functionality to be delivered.

• Split the Merkle test into well-formed crates.
• Prove properties in the individual crates.
• Combine these properties in proofs for the final downstream crate.

If possible, replace some very basic functionality using the standard library, but keep the relevant proofs local to the tree of crates in the Merkle test.

[iamrecursion]

A list of thoughts on this task based off a quick look at the moment:

• Our BinaryHasher trait does not match the Hasher trait in stdlib, so we could factor it out to its own tiny library. The one downside of doing so is that there is nothing useful in that library to prove theorems about. The Hasher trait we have is the same as the stdlib's, but appears not to be used.
• Skyscraper (skyscraper/mod.nr and permute.nr) can be pulled out into a library that makes sense to write proofs in isolation about. Maybe we stick our BinaryHasher in this crate instead of its own, as that is really the only place it is used.
• to_le_bits in utils/bits/mod.nr now can be taken from the standard library, so any theorems specifically about it can go in the corresponding part of the stdlib.
• The same applies to to_le_bytes and from_le_bytes.

All told, we want to factor this into more than one crate and depend on the standard library (with ported theorems) as much as possible! That'll make this a really good showcase of things.