Format docstrings as complete chunks

[akaihola]

Treating large docstrings as separate chunks can cause to weird behavior, e.g. "reformat loops" in which some lines are repeatedly indented and unindented by Darker.

Fixes #240.

• add failing test case
• find multi-line using AST parsing, using e.g.:

  from pathlib import Path
  from tokenize import tokenize, STRING
  from darker.utils import TextDocument
  
  src = TextDocument.from_file(Path("setup.py"))
  readline = (f"{line}\n".encode(src.encoding) for line in src.lines).__next__
  multiline_string_linespans = [
      (t.start[0], t.end[0]) for t in tokenize(readline)
      if t.type == STRING and t.end[0] > t.start[0]
  ]

• if any line in a multi-line string was edited by the user, mark all of the lines edited
• change log
• tests for added functions

[akaihola]

@Carreau, any other ideas for fixing this than doing our own AST parsing to find where possible multi-line docstrings are?

Update: tokenize.tokenize() was sufficient, no need for full AST parsing. I think this is a reasonable solution, what's your opinion?

[akaihola]

@Carreau, this seems to work now, and I found two edge case bugs in my implementation when completing the unit test coverage! I wonder if a second pair of eyes reviewing this would find even more.

I could include this in version 1.4.2 if you have time to review, or alternatively postpone until the 1.5.0 release.

[akaihola]

Moving to version 1.5.0 (1.4.2 was just released).