docs: automated documentation pipeline (per-symbol source → pages)

Replace the hardcoded landing page and minimal reference with a generated
documentation site driven by a single source of truth.

- docs/language/**: one file per symbol (93 keywords/types/builtins/namespace
  methods/operators/annotations), each with front-matter (id, kind, tokens,
  sig, tip) + description + a ```ludic example. Seeded by exploding the former
  inline SECTIONS list; these files are now the source of truth.
- docs/site/: site.json (editable hero/features/showcase/messaging, not
  hardcoded) + snippets/*.ludic (real programs shown on the landing page).
- tools/docgen/gen.py: generates index.html, api.html, ludic-highlight.js and
  symbols.json. The highlighter's symbol tables, hover tips and jump anchors
  are GENERATED from the per-symbol files — add a symbol and it is recognized,
  tipped and linked in every snippet automatically. Python stdlib only.
- tools/docgen/check.py: verifies the pages contract + that no snippet token
  links to a missing reference anchor.
- .forgejo/workflows/docs.yml: rebuilds and publishes to the pages branch on
  every push to main touching the docs sources.

Consumes the new Screen.*/Color.*/named-arg API and the 221-color palette.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

docs/language/random/_section.md

@@ -0,0 +1,7 @@
+---
+id: random
+title: Random
+order: 7
+---
+
+A seeded, deterministic RNG — same seed, same sequence, every run and every platform.