fix(docx): write a paragraph the page reads as markdown as the page sets it

CHANGELOG.md

@@ -8,6 +8,60 @@ follow semantic versioning; release dates are ISO 8601.
 
 ### Public API
 
+- **A DOCX export writes a paragraph the page reads as markdown as the page sets it.** A session
+  reads markdown unless it is told not to (`markdown(false)`). The page then sets a paragraph of
+  plain text holding a mark of emphasis or code: the text its marks style bold or italic, a
+  heading line bold — the first three levels larger — and the marks its parser reads dropped. The
+  export wrote the text as authored, so Word showed `**bold**` with its asterisks and none of the
+  bold, and the report named it.
+  - **The text is read as the page reads it** — line by line, through the page's own parser, a
+    list marker opening a line kept — and written as one run a piece, in the face, family, colour,
+    tracking and size the page sets it in: `**Java**` is a bold run reading `Java`, a heading line
+    bold at its larger size, a line break where the page starts a line. A linked paragraph's
+    pieces stay in one link, and Word's outline lists a heading by the text written. An
+    auto-sized paragraph's pieces are written at its style's size, a heading's at its multiple of
+    it, as its text was.
+  - **The page's own lines decide it.** The pieces are written only where the lines the page laid
+    the paragraph out in hold the pieces' letters in their faces, families, colours and tracking,
+    at their sizes to a hundredth of a point — or, where the page fits the text to a size of its
+    own, at sizes in the same proportion. A session that reads no markdown lays the marks out, and
+    the text is written as it stands, as before; so is text the parser changes nothing of, an
+    underscore inside a word.
+  - **The faces are the page's.** Where the session reads markdown, its parser sets every piece in
+    a face of its own and leaves the paragraph's aside: a bold paragraph's `Senior_Engineer`
+    stands regular on the page, and is written so.
+  - **Every path that writes a paragraph does it:** the body, a table cell, text over the flow,
+    the line an overlay's two sides share, a badge's initials and a page zone's line.
+    - A paragraph composed in a table cell is matched to its lines by its text as the page reads
+      it, where no line carries it as authored, and only to lines that set its pieces so.
+    - A badge's initials are counted as the page sets them: `**JR**` is now a badge's two bold
+      letters. Initials in two faces, `*J*R`, or a heading, are written in the flow, as initials
+      in two runs' faces are.
+  - **A heading written taller than its line is named.** The page sets a markdown heading in a
+    line as tall as the paragraph's own and draws its letters past it; written in that exact line,
+    Word cuts their tops on screen. An auto-sized paragraph's heading, written at a multiple of its
+    style's size, may fit the line the page fits the text to, and is named only where it does not.
+  - **The font table ships the faces the pieces of a paragraph outside table cells and page zones
+    are set in** — the paragraphs it reads. It is written before any paragraph, so it reads them
+    off the text: a session that reads no markdown ships a face it does not use.
+  - **Still written as authored, and named:**
+    - a paragraph whose lines are not read — with no layout, composed in a table cell whose text
+      no line of its table carries so, or a page zone's the layout shows none of — where the note
+      says whether the page reads its marks is not measured;
+    - one the page sets in other letters than its text, as Arabic, which the page shapes before it
+      reads the marks;
+    - text the parser reads into nothing, which the page sets as nothing and which went unnamed:
+      a lone `*`, an empty list item; `***`, a rule; a line set four spaces in, a block of code;
+    - a list's items, as before.
+  - **With no layout**, a paragraph is read line by line for the note too: a list marker opening
+    a line, `* a_b`, is no longer counted as a mark the page drops.
+
+  Across the DOCX fidelity corpus one document's bytes change: `TimelineMinimal`'s open-source
+  project line is written in regular Lato, where Word drew it bold, with `(Open source)` in italic
+  and without the asterisks, as the page sets it; the file ships Lato Italic for it, and the line's
+  note goes: 954 notes, from 955. The other 61 documents are byte-identical. `DocxMarkdown` reads
+  the pieces and compares them with the page's lines, with a unit test.
+
 - **A DOCX page zone's text stands on the page's baseline.** A page zone is written as one line
   of a Word header or footer. The line was Word's single line for its face. Its top sat at the
   zone content's top in a header, and its foot at the content's foot in a footer. So the line's
@@ -142,9 +196,10 @@ follow semantic versioning; release dates are ISO 8601.
   - the marks its parser reads as syntax are dropped, a code span's backticks and a link's
     address among them.
 
-  The DOCX export writes the text as authored, so Word showed `**bold**` with its asterisks and
-  none of the bold, without a note. It still writes it so; the note now says the markdown marks
-  are written as letters:
+  The DOCX export wrote the text as authored, so Word showed `**bold**` with its asterisks and
+  none of the bold, without a note. The note now says the markdown marks are written as letters,
+  where they still are — a paragraph is written as the page sets it wherever the page's lines show
+  how (the entry above):
   - the paragraph's note (`ParagraphNode`);
   - a zone paragraph's, on its `page zone` note;
   - a list's, for its items (`ListNode`).
@@ -159,13 +214,13 @@ follow semantic versioning; release dates are ISO 8601.
   - a marker typed before a list item;
   - runs, which the page never reads.
 
-  Where the lines are not read — with no layout, or composed in a table cell, whose paragraphs are
-  matched to their lines by text and whose lists not at all — the note says whether the page reads
+  Where the lines are not read — with no layout, or composed in a table cell whose paragraph no
+  line its table laid out carries, and a list there at all — the note says whether the page reads
   the marks is not measured, where the page's own parser drops a mark from the text.
-  Across the DOCX fidelity corpus the report names one paragraph: `TimelineMinimal`'s open-source
+  Across the DOCX fidelity corpus the report named one paragraph, `TimelineMinimal`'s open-source
   project line, whose `*(Open source)*` the page sets in italic and Word showed with its
-  asterisks. None of this changes what is written: the 62 documents of the corpus export to the
-  same bytes. In `DocxNodeFieldLedgerTest` a paragraph's `text` and a list's `nestedItems` move
+  asterisks — now written as the page sets it. None of this changed what is written: the 62
+  documents of the corpus export to the same bytes. In `DocxNodeFieldLedgerTest` a paragraph's `text` and a list's `nestedItems` move
   from `WRITTEN` to `REPORTED`, and a list's `items` name it too.
 - **A DOCX export's report names where a page zone's parts stand, and what its paragraphs
   lose.** A page zone is written as one Word line. Word sets its parts one after another from