Two unowned reds on #1160: self-ci Haskell job builds at root with no cabal.project; registry drift on main #1161

Two red checks on #1160 have no tracking issue. Neither is caused by that PR's lock change. This issue is the §5c deferral link for both.

1. Call CI Pipeline / Haskell (and Pipeline report as a consequence)

Latent, newly visible. On main, self-ci.yml dies at startup with zero jobs (run 37437516934). #1160 cures that, and self-ci now creates 16 jobs on the PR head (run 37470262874). That exposes this job for the first time.

• ci-pipeline.yml sets has_haskell by counting *.cabal, stack.yaml and *.hs anywhere in the tree. Here it counts 32 files, all under rhodium-standard-repositories/satellites/cccp/satellites/sanctify-php/.
• The job then runs cabal build all at the repo root. There is no root cabal.project, so cabal stops with Error: [Cabal-7136] There is no <pkgname>.cabal package file or cabal.project file.
• Pipeline report fails because R_HASKELL is failure.

Acceptance criteria

• Either the Haskell job builds each detected package from its own directory, or a root cabal.project lists it, or the satellite is excluded from detection or evicted per [campaign] standards carve-out: evict products/impls, keep the canon (spec + policy + registry + template) #479.
• Call CI Pipeline / Haskell is success or skipped on a self-ci run of main.
• Pipeline report no longer fails on R_HASKELL.

2. Registry + topology in sync

Pre-existing. It fails on main 5751b97 too. The log says DRIFT: .machine_readable/REGISTRY.a2ml is stale — run 'just registry'.

Acceptance criteria

• The registry is regenerated by its own recipe, or the check is retired along with the .a2ml format, whichever the owner rules.
• Registry + topology in sync is green on main.