[Core] Turn on WAL mode for cluster job table

[Michaelvll]

For job table on clusters, we do not turn on WAL mode, which causes errors when user cancels and submits new jobs quickly and frequently:

Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/root/skypilot-runtime/lib/python3.10/site-packages/sky/skylet/log_lib.py", line 448, in tail_logs
    for line in _follow_job_logs(log_file,
  File "/root/skypilot-runtime/lib/python3.10/site-packages/sky/skylet/log_lib.py", line 381, in _follow_job_logs
    status = job_lib.get_status_no_lock(job_id)
  File "/root/skypilot-runtime/lib/python3.10/site-packages/sky/skylet/job_lib.py", line 339, in get_status_no_lock
    rows = _CURSOR.execute('SELECT status FROM jobs WHERE job_id=(?)',
sqlite3.OperationalError: database is locked

[wizenheimer]

Hey @Michaelvll,
Pushed a fix. Please take a look.