Fix edge cases with unicode method names

[HertzDevil]

Fixes #10970, where 😂 is considered an operator:

macro foo(x)
  {% p x %}
end

# Before
foo x.😂(1)  # => x 😂 1
foo x.😂 = 1 # => x 😂= 1

# After
foo x.😂(1)  # => x.😂(1)
foo x.😂 = 1 # => x.😂 = 1

Also fixes some other places where setter methods are currently detected via ascii_letter?:

• Fixes parser failures with the following, as y.😂 is not recognized as a setter target inside a multi-assign:

  class Foo
    property 😂 : Int32?
  end
  
  y = Foo.new
  x, y.😂 = 1, 2

• Fixes formatter failures with the following, as they are not considered setters:

  x._1 = 2
  x.😂 = 1

[Sija]

Shouldn't this be ascii_lowercase? instead of ascii_letter??

[straight-shoota]

Ideally yes. I think the lexer has already ruled out upper case letters as constants at any point where that might be relevant, though. So there should be no practical difference.

[straight-shoota]

Or... maybe not: 😆

"".@FOO # Error: can't infer the type of instance variable '@FOO' of String

[straight-shoota]

So we might want to tighten that down. Probably better in a follow-up. char.ord > 0x9F looks also very permissive.

[HertzDevil]

Uppercase instance variable names are actually allowed in Ruby. Here are the other direct uses of ident_start? in the lexer:

• Reading a symbol literal. :FOO is definitely a valid symbol.
• Reading a percent string literal. This is probably acceptable, because %Q() is a valid string literal.
• Reading a global variable. The only ones that can be declared are lib external variables, and the parser rejects ones starting with an uppercase letter.

Uppercase def names are also allowed in Ruby. Even in Crystal they are allowed for lib funs, so a Call node with an uppercase method name is not necessarily invalid. I don't think we need to do anything here.

[beta-ziliani]

👍