docs: source-align the Echo website in echox

[vishr]

Goal

Make the published Echo documentation accurate against the exact Echo revision it describes while keeping the Astro/Starlight website and all documentation-only files in labstack/echox. Preserve live URLs and important anchors, five locales, cookbook pages, search, Ask AI, and independent website publishing.

This issue was transferred from labstack/echo#3116 after repository ownership was clarified. Echo's API source remains in Echo. Runnable docs examples, reference extraction, authored pages, translations, site code, and deployment belong in echox.

Source revision contract

• Record the Echo commit for each published reference; show it next to generated API facts.
• Compile documentation examples against the recorded Echo source and import the same files into pages.
• Extract facts source can establish: exported config fields and types, deprecations, signatures, and source links. A reviewed baseline must make API drift fail CI with a readable diff.
• Author and review defaults, conditional behavior, security guidance, tutorials, and translations. Extraction alone cannot establish those claims.
• Identify the owning module and version for external middleware.

Implemented in draft PR #447

• Stable and next builds use separate pinned Echo revisions, visible version switches, and separate Pagefind indexes. Existing site URLs remain at the stable root; next lives under /next/.
• Source-backed tables cover all 22 core middleware configs across 20 pages and all five locales. JWT, Prometheus, and OpenTelemetry reference facts come from their pinned module releases.
• Request Logger and Static import the same runnable Go files that the build compiles against the Echo source. These pages correct the drift from Request Logger docs are outdated #439.
• A middleware task page in every locale connects common goals to pages, examples, and follow-on guide tasks.
• CI checks source/API baselines, translated section freshness on the new and example pages, 676 routes, internal links and assets, image text, locale coverage, and representative HTML/asset size budgets. Offline link auditing also checks fragments.
• The existing website deployment remains in echox. Echo cleanup PR revert: keep documentation examples in echox echo#3119 has merged, leaving no documentation-only files in Echo.

Remaining review before publication

1. Review authored behavior, default, and security claims against both pinned Echo revisions; API extraction does not establish runtime behavior.
2. Have native speakers review newly authored Spanish, Japanese, Portuguese, and Chinese task-page wording. Section hashes detect drift but cannot judge translation quality.
3. Measure mobile usability, accessibility, and Core Web Vitals in a browser preview. Current checks cover structural accessibility and static size regressions, not those human/browser measurements.
4. Decide with Echo maintainers whether a small Echo-to-echox cross-repository PR gate is warranted and how it selects an echox ref. The site can already test a proposed Echo checkout with ECHO_SOURCE_DIR.

The gomarkdoc --embed/--check trial and its limits remain in this repository. Package-wide output is too broad for individual middleware pages, and generated fields do not establish runtime defaults or security behavior.

Acceptance criteria

• Existing live URLs and important anchors resolve; all five locales and cookbook pages remain available.
• Removing or renaming a core middleware API field fails the docs compatibility check until the baseline and affected pages are reviewed.
• Cookbook, runnable reference examples, and site build pass against their recorded Echo revisions. Published references visibly identify those revisions.
• Runtime claims, security notes, translations, and external middleware facts have explicit review and ownership paths.
• Reader-task, accessibility, performance, and URL checks are compared with the recorded baseline before publication.

[aldas]

@vishr , please do not move website to echo main repo

[aldas]

it would "pollute" the commit history with "website" related commit noise.

[vishr]

it would "pollute" the commit history with "website" related commit noise.

Noted. Let me look for alternatives.