- lexer.rl: handle CLRF as a line separator

[iliabylich]

Closes https://github.com/whitequark/parser/issues/1020.

3.3.1 :004 > Parser::CurrentRuby.parse("1\r\n2\r\n3").children[2].loc
 => #<Parser::Source::Map::Operator:0x00000001222f0f80 @expression=#<Parser::Source::Range (string) 6...7>, @node=s(:int, 3), @operator=nil>

3.3.1 :005 > Parser::CurrentRuby.parse("1\r\n2\r\n3").children[2].loc.expression.source
 => "3"

A few notes:

1. If \r\n is a line separator parser still emits tNL token with location of the \n character
2. tSTRING_CONTENT tokens now have proper locations, but the content doesn't include \r part of \r\n (because eval(%{"\r\n"}) is just "\n"), so .source of their locations doesn't match string content. I guess it's fine, the same happens with all escape sequences anyway.

If it doesn't break Rubocop's test suite I guess it's safe to merge it as is.

@kddnewton Could you take a look at this please? Does it fix Prism's translator?

[kddnewton]

This gets close, but runs into issues with escaped \r and literal \n then getting grouped, as in:

<<EOS
foo\rbar
baz\r
EOS

(There are regular newline characters after each line.) In this PR, it groups that last \r\n because the gsub is happening after escape sequences are resolved.