Fix NoMethodError in Magik lexer's method_name state

[sebastiaanspeck]

Follow-up to #2318.

state :method_name still called Magik.identifier, a leftover from an earlier revision of that PR that cached regexes as memoized class methods (the style @jneen's review comments on #2318 asked to move away from). That refactor was done for every other regex in the file, but this one call site was missed.

Since RegexLexer evaluates a state's rule arguments eagerly when the state is first loaded, Magik.identifier raises NoMethodError: undefined method 'identifier' for class Rouge::Lexers::Magik as soon as :method_name is entered - i.e. on any input containing _method, not just the new []-method syntax. Verified against a clean checkout of current main (909e8ca):

$ ruby -Ilib -e 'require "rouge"; Rouge::Lexers::Magik.new.lex("_method a.b()\n_endmethod\n").to_a'
NoMethodError: undefined method 'identifier' for class Rouge::Lexers::Magik

This fix replaces the stale class-method reference with the identifier local variable already used everywhere else in the file. It's a one-line change with no other effect: the sample file round-trips through the lexer with no errors, and no other tokenization changes since identifier is the exact same regex Magik.identifier was aliasing.

[jneen]

I'm a little shocked that the tests didn't catch this. Does the visual spec never trigger this state?

[sebastiaanspeck]

I'm a little shocked that the tests didn't catch this. Does the visual spec never trigger this state?

The tests caught it, but it was ignored(?) - https://github.com/rouge-ruby/rouge/actions/runs/32947811182/job/98230454981 and https://github.com/rouge-ruby/rouge/actions/runs/32947811182/job/98230455001

[jneen]

Hm, the action ran but it never made its way to the PR. Incredible.