Simplify switch statement in parse_statements.

[p8]

:on_nl, :on_ignored_nl, :on_comment, :on_embdoc are handled by a single when branch. In this branch there are multiple if branches handling :on_nl and :on_ignored_nl separately from :on_comment and :on_embdoc.

By having separate when branches for :on_nl/:on_ignored_nl and :on_comment/:on_embdoc we can remove the if statements.

[st0012]

I'm not sure if the before/after behaviours are actually the same to be honest.

In this condition, the tk variable is reassigned so the later if statement is comparing the new tk instead of the current one. And this behaviour doesn't seem to be replicated in the updated version.

[p8]

That line would be followed by the following:

unget_tk tk

They cancel each other out, so I removed them.
The next tk would get picked up in the surrounding while loop.

There is some extra processing done at the end of the while loop, but this seems to always get skipped as try_parse_comment is always false and keep_comment is true in both cases.

[p8]

I think the following test case for TestRDocParserRuby reproduces this behaviour.
It is green on both branches.

  def test_parse_regression
    util_parser <<-CLASS
class Foo
  # ignored comment

  ##
  # :method: ghost
  # This is a ghost method

  # This is a regular method
  def regular
  end
end
    CLASS

    tk = @parser.get_tk

    @parser.parse_class @top_level, RDoc::Parser::Ruby::NORMAL, tk, @comment

    foo = @top_level.classes.first
    assert_equal 'Foo', foo.full_name

    ghost = foo.method_list.first
    assert_equal 'Foo#ghost', ghost.full_name
    assert_equal 'This is a ghost method', ghost.comment.to_s

    regular = foo.method_list.last
    assert_equal 'Foo#regular', regular.full_name
    assert_equal 'This is a regular method', regular.comment.to_s
  end

[p8]

I've added a regression test just to make sure.