fix: serve spec-compliant .html.md versions of rendered pages

[keplervital]

Fixes #135.

The proposal on the homepage says:

pages on websites that have information that might be useful for LLMs to read provide a clean markdown version of those pages at the same URL as the original page, but with .md appended. (URLs without file names should append index.html.md instead.)

llmstxt.org itself only follows this for the notebook-built pages. The others 404:

Page	Expected URL	Today
/	/index.html.md	404 (/index.md exists)
/nbdev.html	/nbdev.html.md	404 (/nbdev.md exists)
/domains.html	/domains.html.md	404 (/domains-commonmark.md exists)
/ed.html	/ed.html.md	404 (/ed-commonmark.md exists)
/intro.html	/intro.html.md	200
/core.html	/core.html.md	200

The .html.md files only appear for executed notebooks (via keep-md). The qmd and md pages just get the commonmark render, named {stem}.md, or {stem}-commonmark.md when the source itself is a .md file.

This PR adds a small Quarto post-render script that copies each generated markdown output to its .html.md name, plus the one line in _quarto.yml to enable it.

A few deliberate choices:

• Nothing is removed or renamed. I noticed llms.txt was pointed at /index.md on purpose back in 0536f0c, so /index.md and /ed-commonmark.md keep working exactly as before. The script only adds the .html.md counterparts next to them.
• The script skips files that already exist, so the richer keep-md versions of intro.html.md and core.html.md are never overwritten.
• It is a .ts file rather than Python because nbdev_docs copies sources through _proc before rendering, and that copy drops _-prefixed files and strips the .py extension from Python files. Quarto runs .ts post-render scripts with its bundled Deno, so there are no new dependencies.

Verified locally with nbdev-docs (same as the deploy workflow): _docs now contains index.html.md, nbdev.html.md, domains.html.md and ed.html.md, each byte-identical to its .md source, and all previously working URLs still return 200 when serving _docs.

Known limitation: llmstxt-js.html is a raw HTML resource that never goes through the renderer, so it still has no markdown version.