Signed-off-by: Masatake YAMATO <yamato@redhat.com>

This parser is based on the asciidoc parser and tries to preserve all
features of the regex-based parser (all kinds, full scope,
sectionMarker field, running subparsers for code).

Comparing with the regex based implementation, the rewritten parser
is more than 10 times faster:

  [yamato@dev64]~/var/codebase% CTAGS_EXE=/home/yamato/var/ctags-github/ctags-Regex ./codebase ctags Markdown
   version: 5e6ab4b6
  features: +wildcards +regex +gnulib_regex +iconv +option-directory +xpath +json +interactive +sandbox +packcc +optscript +pcre2
  log: results/5e6ab4b6,Markdown............,..........,time......,default...,2021-12-24-00:27:22.log
  tagsoutput: /dev/null
  cmdline: + /home/yamato/var/ctags-github/ctags-Regex --quiet --options=NONE --sort=no --options=profile.d/maps --totals=yes --languages=Markdown -o - -R code/markdown-corpus code/progit code/selinux-handbook
  686 files, 322854 lines (24506 kB) scanned in 2.0 seconds (12462 kB/s)
  19246 tags added to tag file

  real	0m2.014s
  user	0m1.967s
  sys	0m0.035s
  + set +x
  [yamato@dev64]~/var/codebase% CTAGS_EXE=/home/yamato/var/ctags-github/ctags-C ./codebase ctags Markdown
  version: 258b9cbd
  features: +wildcards +regex +gnulib_regex +iconv +option-directory +xpath +json +interactive +sandbox +packcc +optscript +pcre2
  log: results/258b9cbd,Markdown............,..........,time......,default...,2021-12-24-00:27:34.log
  tagsoutput: /dev/null
  cmdline: + /home/yamato/var/ctags-github/ctags-C --quiet --options=NONE --sort=no --options=profile.d/maps --totals=yes --languages=Markdown -o - -R code/markdown-corpus code/progit code/selinux-handbook
  686 files, 322854 lines (24506 kB) scanned in 0.1 seconds (164033 kB/s)
  19112 tags added to tag file

  real	0m0.190s
  user	0m0.172s
  sys	0m0.015s
  + set +x

@masatake editted the commit log, deleted old implementation, and
implemented code for filling "end:" fields.

Headings defined using underline by either '=' or '-' may contain an
arbitrary amount of '=' or '-' characters which don't have to be equal
to the number of characters on the previous line (which was what the
previous code did).

Special attention has to be paid to code blocks using indentation
(either by 4 or more spaces or a \t character). For instance

    abcd
---

this is a code block followed by underline, while

  abcd
---

is a 'abcd' heading.

The indent can be composed of combination of tabs and spaces and
the previous code didn't take into account combinations like " \t".
Fix that.

We have to check if a sequence of '-'s is followed by a visible character
or not to distinguish

a
- b

which is a list containing 'b' and

a
-

which is a heading.

…#,`,~,-,=

This isn't quite correct check as there may be multiple code spans
on a single line and they can go over multiple lines but at least it
fixes the case the previous regex parser handled.

…hen underlined

- use space between function name and (
- use camel case instead of underline in names