Add argparse-highlight.css with "One Dark" color palette to style
the CLI usage lines with semantic colorization:
- Blue: "usage:" heading
- Purple: program name
- Green: subcommands
- Teal: options (-h, --tree)
- Yellow: metavars

Also strip ANSI codes and disable colors during doc builds to ensure
clean parsing by the argparse lexer.

Add intro paragraphs, wrap argparse directives in `## Command` sections,
and rename generic `## Usage` to descriptive section names for consistent
documentation structure across projects.

Glob patterns like "django-*" in argparse help text trigger RST
"Inline emphasis start-string without end-string" warnings. Add
escape_rst_emphasis() utility that escapes problematic asterisks
(e.g., "django-*" → "django-\*") and apply it in the renderer's
_parse_text() method.

Document sphinx_argparse_neo migration and related fixes for v1.64.0.

Replace .clear() + .extend() with slice assignment (node[:] = children)
to fix mypy errors about Sequence[Node] having no .clear() method and
Node having no .extend() method. This is the idiomatic pattern used in
Sphinx's codebase for modifying node children.

Document when doctests are NOT required:
- Sphinx/docutils visit_*/depart_* methods (tested via integration)
- Sphinx setup() functions (entry points not testable in isolation)
- Complex recursive traversal functions (extract helpers instead)

This clarifies the doctest policy for docs/_ext/ Sphinx extensions
where visitor patterns don't have doctests across docutils, Sphinx,
or CPython's ast.NodeVisitor.

Add working doctests demonstrating:
- Basic section creation from definition nodes
- Page prefix for unique section IDs across docs
- Category-prefixed examples with descriptive IDs/titles

This helper function creates docutils section nodes and is
straightforward to test, unlike complex visitor traversals.

Add working doctests demonstrating:
- Extracting mutex groups from ArgumentParser
- Verifying actions map to same MutuallyExclusiveGroup instance
- Empty mapping for parsers without mutex groups

Uses identity checks (is) instead of set() since dataclasses
aren't hashable by default.

Extract subcommand from parser_info.prog to generate unique section IDs:
- "tmuxp load" -> prefix "load" -> IDs like "load-usage", "load-options"
- "tmuxp" -> no prefix -> IDs like "usage", "options" (backwards compatible)

This prevents duplicate ID warnings when multiple argparse directives
exist on the same documentation page (e.g., documenting subcommands).

Changes:
- Add _extract_id_prefix() static method
- Add id_prefix parameter to render_usage_section()
- Add id_prefix parameter to render_group_section()
- Pass prefix from render() to section methods