fix(deploy): point the staging deployment URL at the host that actually serves it

[frankbria]

Found while verifying the staging deploy after v0.9.2.

The problem

deploy.yml's environment.url pointed at https://dev.codeframeapp.com. That
name resolves to the staging box but has no nginx site block and no
certificate, so the shared VPS answers with another project's cert
(dev.autoauthor.app) and TLS fails:

curl: (60) SSL: no alternative certificate subject name matches
      target host name 'dev.codeframeapp.com'

So the GitHub deployment link looked like staging was down, while it was serving
fine at https://dev.codeframe.sh.

Verified on the box, not inferred

pm2:    codeframe-staging-backend   online   40m
        codeframe-staging-frontend  online   40m
proxy:  nginx active, caddy inactive
sites:  dev.codeframe.sh.conf        ← exists
        dev.codeframeapp.com         ← no site block at all

And the live endpoint reports the commit that was just merged:

{"status":"healthy","commit":"d78281c","deployed_at":"2026-08-11T01:46:43Z"}

/api/v2/settings/keys returns 401, so auth enforcement is working too.

Changes

• deploy.yml — url: https://dev.codeframe.sh, with the reason recorded so
  the next person does not re-derive it
• deploy/README.md — states where staging is and how it is served. The
  working URL appeared nowhere in the repository; only the wrong one did
• legacydocs/nginx-setup-complete.md — SUPERSEDED banner. It documents
  the dead domain and a dead IP (47.88.89.175, replaced by 195.35.14.177 in
  the 2026-07-01 rebuild), so following it wastes real time

Docs and one workflow field only — no application code. actionlint clean, full
suite 6443 passed, 49 skipped.