fix(ops): run heavy commands under a hard memory cap to stop host OOM kills

The host runs with vm.overcommit_memory=0 and no swap, so a process that
grows past free memory makes the kernel OOM-kill across the whole machine
-- the Turbopack build (commit 3d828a61) could take out the database,
nginx or the live release.

Add scripts/with-memory-cap.sh: it moves a command into its own systemd
scope with MemoryMax, so only that cgroup gets OOM-killed (verified: a
Turbopack build died at its 6GB cap, host untouched). Build/analyze/dev/
test*/typecheck now run under explicit caps; ulimit -v is only an explicit
opt-in because it bounds virtual address space per process and 10g/20g both
break V8-based builds. Docker and GitLab builds run in their own isolated
containers with a read-only cgroupfs and opt out explicitly (webpack +
--max-old-space-size stay their bound).

Measured: webpack build peaks ~6.5GB RSS, so 10GB leaves headroom within
the 23.5GB host.

README.md

@@ -969,7 +969,7 @@ branch guard inside `ci-deploy.sh` only accepts `main`/`master`.
 | `pnpm db:bulk`            | Batch-import >50 MB JSON via `scripts/bulk-import-json.ts` |
 | `pnpm db:up` / `pnpm db:down` | Start / stop the `mariadb-turbo` container     |
 | `pnpm gamedata:compress`  | Pre-compress large gamedata JSON to `.gz` (gzip_static) |
-| `pnpm analyze`            | Build + open bundle analyzer                       |
+| `pnpm analyze`            | Build + report per-route bundle sizes              |
 | `pnpm jobs:worker`        | Start background task worker                       |
 | `pnpm biome:check`        | Lint and format code                               |
 
@@ -1098,7 +1098,7 @@ The CMS automatically translates furniture names and descriptions to **13 langua
 pnpm dev          # Start with hot reload
 pnpm typecheck    # Type check all files
 pnpm test         # Run test suite
-pnpm analyze      # Build and analyze bundle sizes
+pnpm analyze      # Build and report per-route bundle sizes
 pnpm biome:check  # Lint and format
 ```