Parser misbehaving in cases like `GenericStruct1<GenericStruct2<int>>`.

[Nyvyme]

Lexer always treats >> as a shift left, but the parser expects two separate >s.

The easiest fix is to check in lexer if character after >> is >, : or ).
This will fix the error in cases like this: test: int = sizeof(GenericStruct1<GenericStruct2<int>>);,
but it will not fix the issue in cases like test: GenericStruct1<GenericStruct2<int>> = ....

Probably need to get rid of <<, >>, <<= and >>= as a token in lexer, but how do we combine them into shls, shrs, etc. when we actually need those? Because parser (I guess we have to do this in parser, I don't actually know) does not know anything about spaces. Need to investigate.

[Nyvyme]

My current plan is to get rid of combined tokens in lexer and modify solc::Token struct to store hint if the character after the token is a whitespace. It should make it possible to combine them into a composite ASTs where it is applicable. Also it will probably resolve this error with nested generics.

Gonna create separate branch for it.