Implement transparent new-types in ssz_derive

[paulhauner]

Issue Addressed

Related to #3625

Proposed Changes

Allows transparent encoding for new-type structs, just like serde does.

/// The encoded/decoded bytes will be identical to a `Vec<u8>`
#[derive(PartialEq, Debug, Encode, Decode)]
#[ssz(struct_behaviour = "transparent")]
struct TransparentStruct {
    inner: Vec<u8>,
}

Additional Info

I added some top-level docs to ssz_derive to make sure we're keeping track of all the attributes somewhere.

I've also moved the tests for the derive macro from eth2_ssz to eth2_ssz_derive. I did this primarily because cargo-udeps can't detect when a dev-dependency is only used in doc-tests and raises a false positive. Moving the tests caused eth2_ssz to be used in tests and therefore udeps was happy again (there's an ignore method for udeps but it was too blunt IMO). It also kinda makes sense to have those tests in the derive crate rather than where they were.

[paulhauner]

FYI @realbigsean here's that SSZ impl we discussed last week ☺️ Does it have all the features you need?

IIRC, you need something like this (which is supported):

/// The encoded/decoded bytes will be identical to a `Vec<u8>`
#[derive(PartialEq, Debug, Encode, Decode)]
#[ssz(struct_behaviour = "transparent")]
struct TransparentStruct {
    serialized_thing:: Vec<u8>,
    #[ssz(skip_serializing, skip_deserializing)]
    local_thing: u8,
}

[realbigsean]

Looks awesome! The new docs are great. I took a stab at also supporting unnamed fields here:
realbigsean@697e190

I included the suggested changes below in that commit ^

[realbigsean]

I don't think these field_names are used

[paulhauner]

Nice catch!

[paulhauner]

I took a stab at also supporting unnamed fields

That was great, I included it in the PR! I made a change in 27a5a36 to support the scenario where there are skipped tuple fields before the non-skipped field.

This is ready for review again ☺️

[paulhauner]

Closed in favor of #3861