The piping is a shell command, but systemd runs ExecStart shell-less,
so the redirect must be embedded in a /bin/bash exec command.

Default environment variables running jupyter will not pick up the
configured conda environment, and actual PySpark jobs will see an
error like "Python in worker has different version 2.7 than that
in driver 3.5, PySpark cannot run with different minor versions".
Since it's tricky to otherwise get the conda environment inline,
it's now a good reason to use the helper script as the ExecStart
command rather than fitting it inline in the .service file. Also,
add a "systemctl status" check at the end since otherwise a failed
exec won't propagate to the init action's error status.