`macro_name_str` should not wrap around full macro path

[c42f]

As of #572 we now have macro_name_str wrapping around the full path of the macro in a macro call, as in

julia> parsestmt(SyntaxNode, "A.b.c\"asdf\"")
SyntaxNode:
[macrocall]
  [macro_name_str]
    [.]
      [.]
        A                                :: Identifier
        b                                :: Identifier
      c                                  :: Identifier
  [string-r]
    "asdf"                               :: String

But this is inconvenient to work with - I think the macro_name_str should just wrap around the identifier c:

julia> parsestmt(SyntaxNode, "A.b.c\"asdf\"")
SyntaxNode:
[macrocall]
  [.]
    [.]
      A                                :: Identifier
      b                                :: Identifier
    [macro_name_str]
      c                                :: Identifier
  [string-r]
    "asdf"                             :: String

In the future, I'd like to enforce that macro_name, macro_name_str and macro_name_cmd only wrap an Identifier (ok, and var), and the only reason we don't have this right now is due to the existence of the syntax @A.b.x

@Keno I assume this is just a bug?

[Keno]

@Keno I assume this is just a bug?

I'm gonna go with yes. I don't think there's any issue with the second representation.

[c42f]

Actually, I think I'd like to bring back the terminals StrMacroName and CmdMacroName because we know those are strictly well behaved. It's just the MacroName terminal which was a bit problematic.

Having tried out the new AST in JuliaLowering, I can say it's definitely simpler and more convenient to work with terminals where possible.

[c42f]

simpler and more convenient to work with terminals

To expand on this - K"macro_name" is awkward because it modifies the semantics of "some part" of the expression inside, and there's simple but nontrivial logic to figure out which part this is. Thus every downstream consumer of expressions needs to repeat that logic.