[ekalinin]

Summary

Fixes #40, fixes #67: a heading that contains a link gets a TOC entry that links to the
wrong place, and its text keeps the raw HTML that GitHub renders for the heading.

Heading	Before	After
## On <sub><sub>[Go to Top](#top)</sub></sub>	* [On <sub><sub><a href="#top">Go to Top</a></sub></sub>](#top)	* [On Go to Top](#on-go-to-top)
## [My issue](https://github.com/ekalinin/github-markdown-toc/issues/67)	* [<a href="https://...">My issue</a>](https://github.com/ekalinin/github-markdown-toc/issues/67)	* [My issue](#my-issue)
## See [docs](#top) here	* [See <a href="#top">docs</a> here](#top)	* [See docs here](#see-docs-here)
## A  B	* [A <a ...><img ...></a> B](x.png)	* [A B](#a--b)
## Linux vs. Samba ***permissions***	* [Linux vs. Samba <em><strong>permissions</strong></em>](#linux-vs-samba-permissions)	* [Linux vs. Samba permissions](#linux-vs-samba-permissions)

• gh-md-toc (gh_toc_grab):
  • The link was the first href of the heading line, so a link in the heading text came
    before the heading's own anchor. href_regex now matches the anchor element
    (class="anchor"), and href is taken from it, in both awk branches (default and
    OS/390).
  • All HTML tags are removed from the heading text in common_awk_script; the text
    inside them is kept. HTML entities stay as GitHub renders them
    (Tom & Jerry <3). It is done on text in awk, not with sed over the line,
    which would also remove the anchor element.
  • Steps 4 and 5 of the comment at the top of the file are updated to the current
    expressions.
• tests/tests.bats: a new test on a new fixture test_headingmarkup.md. The Repairing setext heading support, certain uses of backticks after GitHub API format changes #145 test
  now expects * [Title two](#title-two) instead of * [<em>Title two</em>](#title-two),
  and the same line is updated in the TOC of test_setextwithformatting.md, because that
  test runs with --hide-footer (--insert) and rewrites the fixture.
• openspec: the fix-heading-markup change (proposal, design, specs, tasks), not
  archived yet.

Output change: entries of headings with inline formatting or inline HTML (*em*,
**strong**, <sub>, links, images) lose the tags, so a TOC that --insert put into
such a file changes on the next run. Entries of headings without tags do not change.

Not included: decoding HTML entities, collapsing the spaces left by a removed element
(A B, as in GitHub's own aria-label), keeping backticks in entries (#153), the
separate OS/390 text regex.

Testing

• The new test and the updated Repairing setext heading support, certain uses of backticks after GitHub API format changes #145 test fail without the change; all 50 tests pass with
  it (macOS, BSD awk and sed).
• The output for README.md, every other fixture and the four remote pages from the
  remote-toc spec is byte-for-byte the same as with the script from master.
• make lint passes.