Implement PrimaryKey for &Addr by removing unused parse_key

[webmaster128]

The trait PrimaryKey puts a lot of requirements on the type implementing it. It requires zero-copy serialization and zero-copy deserialization.

However, deserializing the bytes into the key type is never used. If we remove it, we can easily implemement PrimaryKey for &Addr.

---

Do not merge as is. Target branch is a copy of the fork from #260.

[maurolacy]

I like the simplicity of this. I think parse_key is not used because we're implementing parsing of keys when we need it using custom helper functions.

[ethanfrey]

I think parse_key is not used because we're implementing parsing of keys when we need it using custom helper functions.

There is an open issue to make use of parse_key: #198 and thus avoid the helper functions.

It does simplify things to remove the call. Let's discuss whether we want to make #198 impossible first. If that is fine, let's do this approach. Otherwise, let's do the &K approach, which is quite nice

[maurolacy]

I think parse_key is not used because we're implementing parsing of keys when we need it using custom helper functions.

There is an open issue to make use of parse_key: #198 and thus avoid the helper functions.

My understanding is that #198 can be done independently of parse_key (at least, implicitly, i.e. automatically). Anyway, consolidating parse_key functionality wouldn't be bad.

It does simplify things to remove the call. Let's discuss whether we want to make #198 impossible first. If that is fine, let's do this approach. Otherwise, let's do the &K approach, which is quite nice

I agree #265 looks like the way to go. If we're OK with tolerating some syntax issues / constructs (mostly in test code), I would say, let's go for it.

[ethanfrey]

I'm happy to merge this.

Let's add the parse_key when we want to take on #198 and we can do #265 at the same time, with a real benefit to such work

[ethanfrey]

Good solution.

[ethanfrey]

Part was merged in #267

The rest cherry-picked into #260

All code used, just not this PR