[Parse] Simplify parsing name for function declarations

[rintaro]

Previously, when parsing func ==<T>(
The parser used to consume ==< as an operator, then mark it as split tokens.
In this PR, consume 2 characters (excepting <) as tok::identifier tok::oper_binary_spaced, then reset Lexer position at <.

@benlangmuir This PR changes the kind of the token into tok::identifier which affects SourceKit doc-info results. Is it OK? I want to make it so because libSyntax expects tok::identifier token at this position.

[rintaro]

@swift-ci Please smoke test

[benlangmuir]

Does this really only affect doc-info, or is that just where it happened to be tested? Wouldn't this affect all the other syntactic sourcekit requests?

Also, to clarify: this isn't just changing the token kind for ==, but for any operator declaration, right? That doesn't seem correct to me. We don't treat a reference to an operator as an identifier in SK, so why would we should the decl as an identifier? + @akyrtzi as well.

[akyrtzi]

It looks like for the syntax-map (sourcekitd-test -req=syntax-map /path/to/test.swift) operators are not included at all (both in the decl or in expressions). We could preserve current behavior for sourcekitd by checking if the 'identifier' is an operator and avoid reporting it if it is.

[rintaro]

@swift-ci Please smoke test

[rintaro]

OK, I changed it to tok::oper_binary_spaced.
We can make libSyntax to accept operator tokens at this position.