chore(ci): track staging for the Dagster+ prod code location

[jirhiker]

CD_dagster_prod.yml triggered on pushes to production, which is still on
1.2.2 and has no automated_ingestion/ directory — so it could never fire.
Every deploy so far has been a branch deployment from a pull request, leaving
the Dagster+ prod code location empty. That is why the Dagster+ UI sits on
"Waiting for GitHub Actions build to start".

What changes

One line: branches: [production] → branches: [staging].

"prod" names the Dagster+ deployment, not this repository's production
branch. Tracking staging matches how ocotillo-api-staging already behaves
and keeps the code location current with where the work is.

Why not leave it on production

The original reasoning — a pipeline writing to the production database should
follow the release flow — still holds. It is just premature: the pipeline has
no database or vendor credentials yet, so a deploy publishes code, not data,
and the worst it can do is fail to materialize.

The comment now records when to act on it: move the trigger back to
production in the same change that first points the pipeline at the
production database. Doing it then keeps the decision attached to the risk
that motivates it.

After merge

The merge itself is a push to staging touching a path in the filter, so it
should deploy the code location on its own. ingestion_heartbeat is the asset
worth materializing to confirm — it touches nothing, so success means the
location loads and runs.

🤖 Generated with Claude Code

[github-actions]

Coverage

✅ 79.26% total — gate is 75%.