docs: teach the docs native HTML and CSS, drop html-webpack-plugin and the CSS loaders - #8355

The splash example bundled one JavaScript file and asked the reader to
write the page by hand, pointing a <script> tag at dist/bundle.js. Webpack
handles all three now, so the example starts from src/index.html: the page
links a stylesheet and a script, entry points at the page, and the emitted
document is shown next to the config with its URLs rewritten to the hashed
assets. Verified against webpack 5.110.3 — an HTML entry turns on the html
and css support on its own, and htmlFilename keeps the page's name out of
the contenthash pattern.

The migration guide's companion-plugin table now covers the popular
html-webpack-plugin ecosystem rather than five entries, and a new section
maps its six hooks onto HtmlModulesPlugin's, with a plugin rewritten
through transformTags and injectTags as the worked example.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NuWDbcF2jvq7q1r8aEY5V9

… docs

Webpack handles CSS itself, so the three packages are no longer named
outside the migration guide, and their pages are gone: css-loader and
style-loader join the excluded loaders, MiniCssExtractPlugin the excluded
plugins, and /loaders/css-loader, /loaders/style-loader and
/plugins/mini-css-extract-plugin redirect to the Native CSS guide. The
exclusion entries carry the webpack/ prefix the repositories actually
have; the older webpack-contrib/ spellings around them no longer match
anything the fetcher lists.

The HMR guide stops installing style-loader and css-loader for its
stylesheet section: experiments.css defaults to 'auto', so importing a
stylesheet needs no rule at all — webpack bundles it and links it into the
page, verified against webpack 5.110.3. The remaining mentions were
reference asides (a filter option, an ident-hash comment, an exports
convention, the CSS extraction bullet) and now describe the behavior
without naming a package, linking the guide where a reader needs more.

Left alone: the Native CSS migration guide, which has to name what it
migrates from, and the release posts and the v1-to-v2 migration guide,
where the loaders are part of the record.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NuWDbcF2jvq7q1r8aEY5V9

webpack/webpack#22008 gives a page's extracted scripts the entry's
filename now that the page emits no JavaScript of its own, so the guides
stop showing generated identifiers. output-management and development list
main.bundle.js and main1.bundle.js for the two scripts the page pulls in,
the PWA listing follows in the same format, and caching is back to
main.<contenthash>.js beside runtime and vendors — its note about generated
chunk names is gone and the prose names the main bundle again.

Every listing was regenerated against that branch (84c1155): a page with
two scripts emits main.bundle.js and main1.bundle.js, and the caching setup
emits main.<hash>.js, vendors.<hash>.js and runtime.<hash>.js.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NuWDbcF2jvq7q1r8aEY5V9

The two owner-rename tests used css-loader as their example, and this
branch stops building a css-loader page, so the "has a page" assertion
no longer holds. postcss-loader is the same shape — a webpack-owned
loader the site still builds a page for — and keeps both tests testing
what they were written for.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NuWDbcF2jvq7q1r8aEY5V9