fix(queen): drop the duplicate npm ci that Nixpacks cannot delete

[gHashTag]

The queen has never deployed

All 13 deployments of the trinity service (project overflowing-renewal, env production) failed. Zero successes.

The first ten failed on the root Dockerfile, which runs find /app/.zig-cache in a fresh runtime stage where /app does not exist — and, having no COPY --from=builder, never ran the builder stage at all.

With the service pointed at apps/queen-web and Nixpacks, the failure moved here:

npm error EBUSY: resource busy or locked, rmdir '/app/node_modules/.cache'
Build Failed: ... "npm ci && npm run build" ... exit code: 240

Cause

Nixpacks mounts a build cache at /app/node_modules/.cache and already runs npm ci in its own install phase:

║ install    │ npm ci                        ║
║ build      │ npm ci && npm run build       ║

Our buildCommand ran npm ci a second time. npm ci deletes node_modules wholesale before installing, and it cannot rmdir the mounted cache directory — so it aborts with EBUSY.

Fix

buildCommand: "npm ci && npm run build" → "npm run build". One line.

Verification

The application was never the problem. Against this exact tree, locally:

command	result
npm ci	73 packages, exit 0
npm run build	83 modules transformed, 253.49 kB bundle, exit 0

One thing worth knowing

This file overrides service settings made through the API or dashboard. Setting buildCommand on the service instance had no effect while this line said otherwise — the build kept using the value from here. That cost two deployments to establish.