rustla2-api: glog writes unbounded log files to /tmp (no retention)

Nobody has claimed this yet.

`rustla2-api` logs via glog's default file sink, which writes into `/tmp`
inside the container. glog opens a **new file per process start** and never
removes old ones, so a long-running deployment accumulates log files
indefinitely.

Observed on a long-lived container: ~50 files totalling ~6.7 GB, the oldest
dating to early 2024. Individual files reached 1.8 GB.

This is invisible in normal operation because a redeploy creates a fresh
container with an empty writable layer, discarding everything — then it
accumulates again from zero.

### Note on `--max_log_size`

Setting `--max_log_size` is **not** sufficient. That flag only rolls over to a
new file once the current one hits the cap; it does not delete previous files.
It would increase the file count without bounding total disk use.

### Suggested fix

Run with `--logtostderr` (or env `GLOG_logtostderr=1`) so logs go to stderr and
are captured by the container runtime, where they can be bounded by the log
driver's `max-size` / `max-file` options. This also makes `docker logs` useful
for this service, which it currently isn't.

If file logging is preferred instead, `--log_dir` should point at a mounted
volume with a logrotate policy attached.

Start by locating the rustla2-api container launch configuration and its current glog setup. Verify how logs are written today, then make the service use stderr or a mounted volume with rotation; done means old files no longer accumulate unboundedly and container log collection is useful.

Written by the indexing model from the issue text.