Add optional blank line preservation (MD_FLAG_PRESERVEBLANKLINES)

[eszlamczyk]

Closes #384.

Motivation

When rendering from the MD4C callback stream (e.g. building a native view tree for a WYSIWYG-like editor), the number of blank lines the author typed between blocks is meaningful. CommonMark collapses any run of blank lines into a single block boundary, and MD4C discards this information in the process so it can't be recovered downstream without pre-processing the source. This adds an opt-in way to get that information from the parser.

What this does

Adds the flag MD_FLAG_PRESERVEBLANKLINES. When enabled, each run of blank lines separating two blocks is reported as a single new block MD_BLOCK_BLANK, whose detail MD_BLOCK_BLANK_DETAIL carries the number of blank lines:

  typedef struct MD_BLOCK_BLANK_DETAIL {
      unsigned count;   /* Count of blank lines forming the block separation */
  } MD_BLOCK_BLANK_DETAIL;

Following the discussion in #384, the run is coalesced into one block (the count is stored once, not one block per line), and all blank lines are reported - i.e. the arser reports "this block separation is made of N blank lines" and leaves the rendering decision to the application.

With the flag off, behaviour and output are byte-for-byte unchanged.

Semantics

• One MD_BLOCK_BLANK per run; no MD_TEXT is sent.
• Blank lines attach to the innermost open container and flush at the next block boundary, so runs split correctly across edges: blanks inside a block quote are reported inside it, blanks after it at the outer level (and vice versa).
• Only genuinely empty lines are counted. Blank lines inside code blocks are verbatim content, and a few non-blank lines MD4C internally treats as blank (a closing code fence, an admonition tag) are excluded.

md2html gains --fpreserve-blank-lines; its renderer emits count non-breaking-space paragraphs so the count is visible/testable.

Note: Blank lines trailing a list's last item (before a dedented paragraph that ends the list) currently attach inside that last <li>, consistent with the "innermost open container" rule. Happy to tweak those semantics, if you'd prefer. Documented with a test in test/coverage.txt

Tests & docs

• New test/spec-preserve-blank-lines.txt (10 examples, doubling as docs) plus harder paths in test/coverage.txt (closing-fence exclusion, HTML-end blank, loose lists, trailing-in-item, nested containers, whitespace-only lines).
• Full suite green; -Werror Debug + Release clean. CRLF verified manually but not added as a spec test (the .txt files are LF-only).
• README, CHANGELOG, md2html man page, and md4c.h doc comments updated.

[mity]

Looks as the right approach to me. There are some requested changes though.

[codecov]

Codecov Report

❌ Patch coverage is 76.92308% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 78.87%. Comparing base (9de89d8) to head (5b5322b).
⚠️ Report is 1 commits behind head on master.

Files with missing lines	Patch %	Lines
src/md4c.c	75.00%	1 Missing and 4 partials ⚠️
src/md4c-html.c	80.00%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #385      +/-   ##
==========================================
+ Coverage   78.86%   78.87%   +0.01%     
==========================================
  Files           5        5              
  Lines        3794     3820      +26     
  Branches     1271     1279       +8     
==========================================
+ Hits         2992     3013      +21     
- Misses        297      298       +1     
- Partials      505      509       +4     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.