Hand a provisioned worker its token in the environment, not argv

The generated batch script exported nothing and passed `--token`, which
any user on a shared compute node reads from `ps`. It now exports
FLUKSIO_WORKER_TOKEN after `prerun`, which the worker already reads.

backend/fluksio/flow/provision.py

@@ -231,7 +231,6 @@ class SlurmProvisioner:
         worker = [
             "exec fluksio-worker",
             f"--url {self.engine_url}",
-            f"--token {token}",
             f"--cpus {profile.cpus}",
             f"--gpus {profile.gpus}",
             f"--max-idle {self.max_idle_s:.0f}",
@@ -248,6 +247,10 @@ class SlurmProvisioner:
             *[f"#SBATCH {option}" for option in profile.sbatch],
             "",
             *profile.prerun,
+            # Not `--token`: argv is in `ps` for every user of a shared node,
+            # the environment only for the job's owner. Exported after
+            # `prerun`, so nothing it starts inherits the token.
+            f"export FLUKSIO_WORKER_TOKEN={token}",
             " ".join(worker),
             "",
         ]

backend/tests/flow/test_provision.py

@@ -87,6 +87,9 @@ def test_the_script_starts_a_worker_that_dials_back_and_stops_when_done(monkeypa
     assert "--labels gpu" in script
     # Without this the job holds its allocation until the walltime.
     assert "--max-idle 120" in script
+    # The token travels in the environment: argv is in `ps` for anyone.
+    assert "--token" not in script
+    assert "\nexport FLUKSIO_WORKER_TOKEN=ey" in script
 
 
 def test_one_job_per_profile_however_often_it_is_asked_for(monkeypatch):

docs/code/workers.md

@@ -240,6 +240,9 @@ When a node needs a machine nothing attached can give, and a profile fits, the
 engine `sbatch`es one over the system `ssh`, and the run waits meanwhile, saying
 so. `prerun` owns the environment: a `module load`, or a venv with
 `fluksio-worker` already in it. The generated script runs no `pip install`.
+It hands the worker a token of its own, good for 24 hours, as
+`FLUKSIO_WORKER_TOKEN` rather than `--token`, so other users of a shared
+compute node cannot read it from `ps`.
 
 One outstanding request per profile, however often it is asked for. A job that
 never attaches within `provision_timeout_s` is `scancel`led, as is anything