publish-governed-sqlite-registry tutorial has no CI replay, and its exclusion comment claims otherwise

[jeremi]

Two related gaps

1. The exclusion comment overstates what is proven

docs/site/scripts/check-evidence-tutorials.sh:138 excludes the Relay tutorial with this reason:

publish-governed-sqlite-registry   # Relay V2 journey; replayed by the Relay product and real-process acceptance gates

Those gates do not replay this page. products/relay-v2/acceptance/business-registry/ is a
materially different and much larger synthetic contract (two resources, list/read/search, protected
access profiles, geo search, pagination). It exercises the Relay engine, but nothing extracts this
tutorial's fenced blocks, runs its commands, or compares its printed output. No script in the repo
does. The comment reads as coverage, and it is not.

2. The page therefore has no replay at all

docs/site/scripts/check-tutorial.sh targets only first-run-with-solmara-lab.mdx.
check-evidence-tutorials.sh skips this page by the exclusion above. In ci.yml the docs job
runs only dry-run tutorial checks. So publish-governed-sqlite-registry.mdx is not replayed in CI
in any mode: nothing checks that its commands still run, let alone that its printed values still
match real output.

The tutorial installs from the mutable releases/latest URL, so its shown output can silently
diverge from what a reader gets on any new release.

Why this is worth closing properly

The one tutorial in this tree that does show exact digests, publish-and-consume-discovery-index,
earns it: docs/site/scripts/check-discovery-tutorial.sh greps a real CI transcript for those exact
strings, so drift fails the build and someone updates the page. That is the standard worth matching.
Everywhere else the convention is angle-bracket placeholders precisely because no gate asserts the
bytes.

Suggested direction

Either give this page a real replay gate in the style of check-discovery-tutorial.sh, which would
let it show exact values honestly, or leave it on placeholders and correct the exclusion comment so
it states the truth: that the page is not replayed.

Provenance

Found while reviewing #785. That PR applies the placeholder convention to the drift-prone values on
the page, which removes the immediate misleading-output problem. It does not add a replay gate and
does not touch the exclusion comment.