v0.17.5

Fixed

• docrev import manuscript.docx keeps the reviewers' work. Starting a project from a Word file read only its plain text: comments, tracked changes, images and equations were lost, and a paper whose headings were not exactly "Introduction", "Methods" and the like came out as one content.md. The import now reads the document as docrev sync does: comments and tracked changes become CriticMarkup, images go to figures/media/, equations become LaTeX, and the title and authors of the Title and Author paragraphs go into rev.yaml. It splits the text at the document's top heading level, with the text before the first heading in preamble.md; a document without Word heading styles is split at paragraphs that consist of a standard section name, as before. On the test manuscript (298 comments) the import keeps every comment, the tracked changes and 7 images in 9 section files, where it used to write one file with none of them.
• Tracked changes nested inside other changes or comments come through intact. Word lets a comment end inside a deletion and a deletion fall inside an earlier insertion; reading pandoc's output text with patterns left such spans as raw [...]{.deletion author=...} in the Markdown. The changes are now rewritten in pandoc's document tree, so the nesting resolves: a comment marker inside a change disappears with the comment, and a change inside the opposite change (text inserted and then deleted) is dropped as having no effect. This applies to sync, import, status and comments.
• Comments land on the text they were made on. Anchors were searched in a lower-cased copy of the Markdown, and a character whose lower case is longer (Turkish "İ" becomes "i" plus a combining dot) moved every later comment one character into its anchor, so a comment on "Discussion" came out as D{>>...<<}[iscussion ]{.mark} and could take the + of a closing ++} with it. Matches found after ignoring whitespace or CriticMarkup were also placed at their offsets in the altered copy. All matches are now mapped back to the Markdown itself; on the test manuscript all 280 comments found by their anchor text land on it, where 262 of them were off before. A comment whose anchor spans tracked changes is placed at the start of the span without a highlight, so it never breaks the change markup.
• A comment with no highlighted text keeps the space after it. Alps.{>>R2: How were the plots placed?<<} Each plot built as "Alps.Each plot" in Word: the comment borrows the next word as its anchor, and the space before that word was dropped with it. The space now stays outside the comment's range, so the DOCX reads "Alps. Each plot".