…e Search Console

  Docs (docs.feldera.com):
  - Set `trailingSlash: true` to match the host, which serves canonical
    pages with a trailing slash - this avoids unnecessary redirects for
    links, canonical tags and sitemap entries, generated by Docusaurus.
  - Sitemap: emit `lastmod` (the only freshness signal search engines
    honor) and drop utility routes and the raw OpenAPI spec from it.
  - Add `static/robots.txt` (allow all + sitemap pointer).
  - Fix internal doc links to use explicit `.md`/`.mdx` targets so they
    resolve to the canonical URL instead of a redirect hop.
  - De-slash blog/paper links (what-is-feldera, literature/papers x2,
    fraud_detection) to drop the 308 hop to the canonical no-slash URL.
  - Add three client-redirects for real 404s (eks/ingress, k3d, kafka/
    debezium) so old indexed URLs land on live pages.
  - Mark the sandbox "Run in sandbox" link `nofollow`: it deep-links to
    `try.feldera.com/create?code=...`, and following it floods Search
    Console with duplicate/redirect entries for un-indexable app URLs.
  - Remove the unused markdown-page.md example stub.

  Pipeline-manager:
  - Serve `GET /robots.txt` with a blanket `Disallow: /`. The web console
    is a client-side SPA, so per-page robots hints never reach crawlers; a
    root disallow is the only reliable way to keep app URLs (the sandbox's
    `/create?...` deep-links) out of search indexes. Covers every
    deployment, which is the desired default. Regression test included.
  - Exclude the test-only `POST /pipelines/{name}/testing` endpoint from
    the public OpenAPI spec (dropped from `paths()`, route kept for the
    test harness); regenerate openapi.json. Stops docs from publishing —
    and Google from indexing — a URL that was never a supported API.

Signed-off-by: Karakatiza666 <bulakh.96@gmail.com>