fix(ci): move docker-e2e test scratch off tmpfs to disk (#702)

The nightly Docker Container E2E job writes every fork's image and video
scratch under /tmp, which docker-compose.test.yml mounted as a 2g tmpfs.
On the 7GB runner that scratch either exhausted RAM (the runner was killed
mid-run, exit 137) or, once the tmpfs was sized to cap RAM in an earlier
attempt, hit ENOSPC and failed ~1100 tests. Drop the test-unit tmpfs mounts
so scratch lands on the container's disk-backed layer, which the job already
frees to ~110G. Keep VITEST_MAX_FORKS=1 for the memory ceiling; Postgres and
Redis keep their small tmpfs data mounts.

docker/docker-compose.test.yml

@@ -62,17 +62,16 @@ services:
       - SYNC_WAIT_MS=120000
       - VITEST_TEST_TIMEOUT=180000
       - VITEST_HOOK_TIMEOUT=120000
-      # /tmp is tmpfs (RAM-backed) below, so each fork's scratch files count
-      # against RAM alongside the app plus Postgres and Redis. On a 7GB CI
-      # runner the default fork-per-core parallelism has correlated with the
-      # runner being killed mid-run (exit 137). One fork bounds the peak; the
-      # 90-minute job budget absorbs the slower serial run.
+      # One vitest fork bounds peak memory on a 7GB CI runner; the 90-minute
+      # job budget absorbs the slower serial run.
       - VITEST_MAX_FORKS=1
-    tmpfs:
-      # Cap the RAM-backed scratch mounts so a burst of large temp files (a
-      # multi-megapixel encode, a video transcode) cannot exhaust runner RAM.
-      - /tmp/test-workspace:size=1g
-      - /tmp:size=2g
+    # No tmpfs: the tools write multi-megapixel encodes and video transcodes to
+    # WORKSPACE_PATH and /tmp, and a RAM-backed mount made that scratch count
+    # against the 7GB runner alongside the app, Postgres, and Redis, which
+    # correlated with the runner being killed mid-run (exit 137). Sizing the
+    # tmpfs instead just moved the failure to ENOSPC. Let scratch land on the
+    # container's disk-backed layer, which has ample room once the job frees the
+    # preinstalled toolchains.
     depends_on:
       postgres:
         condition: service_healthy