Fix StrNode#last_line and StrNode#line_count for heredocs

[fatkodima]

Parser's heredoc format:

(str "foo\nbar\n")
'<<HERE␊foo␊bar␊HERE'
 ~~~~~~ expression
        ~~~~~~~~ heredoc_body
                ~~~~ heredoc_end

Right now, heredocs are not handled correctly, imo. Every updated method in this PR in the original implementation only deals with expression part. And so returns incorrect values (look at tests for expected behavior).

Rubocop needs a couple of changes (2-3 very small) due to new behavior to let the CI pass. But this gem depends on rubocop, and rubocop depends on this gem, so I need and advise on how to proceed and in what order.

[marcandre]

Heredocs are complicated. The notion of 'source' of a heredoc is strange as the source is disconnected.

It's important to realize how disconnected it is...

foo(<<~RUBY, other, arguments(:here), not(related_to_here_doc))
  inside heredoc
RUBY

I don't think that including , other, arguments(:here), not(related_to_here_doc))in the source will be of much help to deal with heredocs...

[fatkodima]

Yeah, agreed.
But what about #first_line, #last_line and (not present in this pr) #line_count methods?
I'm working on a pr into main repository, where I need to get the number of lines in heredoc, so specially interested in #line_count method.

[fatkodima]

Feel free to ping me if there would be any interest in this.

[marcandre]

Thanks for the ping, sorry I didn't answer before.

I'm 👍 on #first_line, #last_line, and (not present in this pr) #line_count`, and anything else that can be useful to you or others and is well defined :-)
Please have tests for complex cases (only), e.g.

x = foo(<<~FIRST, <<~SECOND)
  part of first heredoc
FIRST
  this is
  part of second heredoc
SECOND

[marcandre]

We can consider line_range (instead?), and maybe they should be named body_line_range or heredoc_line_range to underline that it is the range of the text itself, not of the begin/end markers

[fatkodima]

On master branch, for heredocs:
StrNode#first_line works as expected,
fixed #last_line to return the correct line (of the ending marker) - on master always returns the line of first marker
fixed #line_count to return the correct line count (2 markers + of content) - on master always returns 1 as lines in first marker