LLMs-code.txt

[pwizla]

This PR introduces a new consolidated llms-code.txt file with all examples found in the documentation. To do so, it:

• Extracts all code examples from MD/MDX pages, grouped by page and by section, with a consistent block per example: Description, Source (with
  optional anchor), Language, File, and the fenced code. Language variants (e.g., JS/TS) are grouped under the same example with a ‘---’ divider.
• Adds a validator to ensure structure, recognized languages, fence consistency, and (optionally) file existence and section anchor presence
  (including custom {#custom-id}).
• Supports generating across all docs with an “all docs” mode and optional include/exclude filters for scoping; also supports stdout previews to
  review format quickly.
• Integrates generation and validation into the dev/build flows and provides a single “generate+validate” command for local runs and future CI use.
• Leaves existing llms.txt and llms-full.txt untouched; this is an additive feature focused on code examples.

Usage hints:

• Generate and validate all docs: from docusaurus/, run npm run llms:generate-and-validate
• Validate only: npm run validate:llms-code