Improve grammar after the introduction of case-in pattern matching

[aibaars]

This pull request implements some simplifications and minor fixes .

• The "value" expression of case value in ... is no longer optional
• The spurious singleton 'choice' in the case_match rule is removed
• Constants in patterns now use the same node names as normal constants (ie constant and scope_expression)
• The various forms of * and ** in array and hash patterns now use the same node names as the ones used in formal parameter lists (ie splat_parameter, hash_splat_parameter)
• The **nil token is now named hash_splat_nil and is used in patterns as well as formal parameter lists (used to be missing)
• pattern_pair is renamed to keyword_pattern to match the naming of keyword_parameter
• pattern_variable was removed, it is a simple identifier and there is no real need for an additional wrapper node .
• Ranges in patterns now use the same node names as normal range (ie range)

Checklist:

• All tests pass in CI.
• There are sufficient tests for the new fix/feature.
• Grammar rules have not been renamed unless absolutely necessary.
• The conflicts section hasn't grown too much.
• The parser size hasn't grown too much (check the value of STATE_COUNT in src/parser.c).

[nickrolfe]

I don't think this change is correct, since the value does appear to be optional. MRI accepts code like this:

def f(a, b)
  case
  when a > 1
    puts 'foo'
  when b == 7
    puts 'bar'
  else
    puts 'baz'
  end
end

Here is an example I found in the wild.

Compare parse.y line 3134 with line 3118.

[aibaars]

That case is covered by the case rule. The expression case in pattern then ... is not valid. The missing value is only allowed in combination with when branches.

[nickrolfe]

Ah, sorry, that makes sense.

[nickrolfe]

I went though each commit one-by-one, and they all look sensible to me.