Support flattening attributes

[ecton]

Similar to #4, I'm adding a new attribute for ViewSchema and want to share the view() identifier name, so that all the parameters just live on the same level, but I can avoid repeating the shared attributes on both types. Using the same example from #4, what I would hope is for:

struct Attr {
   #[attribute(flatten)]
   shared: SharedAttributes
}

struct SharedAttributes {
    shared_key: String
}

Should parse

#[attr(shared_key = "hi")]

[ModProg]

Yeah that sounds like a reasonable request. Initial idea would be that the AttributeParser gets a parse_one function, that parses one key(= value)? pair.

So in the attribute parsing step every field could do one parse. Designed right this could open the door for more advanced parsing e.g. nested structs as well.

[ecton]

I want to point out that while I still think it's a neat request, I've decided to keep the derive's using separate attributes for now -- it just felt a little weird thinking about how you'd split the implementations if you didn't want to ship the ViewSchema to the client, since it's primarily a server/local trait despite being defined in core.

So this feature isn't actually something I'd be using currently.