Rust API for BLS signatures and EVM precompiles

[Vindaar]

And here we are finally for the first part of the Rust API, namely BLS signatures.

This is still a draft, because there's again a few questions about how this should ideally be wrapped. Currently the code simply uses type aliases similar to the Go API. This is "fine", but leads to the user having to write code like:

        let mut pub_key = MaybeUninit::<EthBlsPubKey>::uninit();
        let status = deserialize_pubkey_compressed(pub_key.as_mut_ptr(), inp.as_ref());
        let pub_key = unsafe { pub_key.assume_init() };

which imo is rather ugly.

2 alternatives come to mind:

• at least for these types simply have deserialize_*_compressed (and friends) return the deserialized type directly. That way we can hide the unsafe code in the fake constructor.
• wrap the C types in a Rust struct and abstract it away that way. This approach will just mean we have to do more work on API implementation side, but at least we get shorter function names (because we can then "overload" names like deserialize) and we shouldn't suffer from copies, because we can of course just pass the pointers of the underlying object to C.

Another point:
Funnily (or ironically), while we can nicely turn the nested messages in batch_verify etc. into a bunch of CttSpans, we still have to make a copy of the data due to Rusts extreme stance on pointer mutability. Given that the messages argument is just a reference (wouldn't make sense to force it to be &mut), we are not allowed to turn them into a CttSpan (or ctt_span to be exact), because it is defined as having a byte* field and not a const byte* field. I'm a bit perplexed that we cannot even cast an immutable to a mutable pointer inside of an unsafe block. At least that's my understanding so far.
I'm not sure if we want to restrict us so far as to have the ctt_span have a const byte* field, because there should be valid reasons to use the type for mutation as well? We could of course have a ctt_span_const, but that's also a bit ridiculous.

Anyway, while the tests are not pretty due to so much error checking and MaybeUninit work, at least they all pass.

[Vindaar]

Just finished porting over the EVM precompiles as well. Decided to just do it in the same PR now.