#827 CLI hwe terminate/reset fully broken: UPDATE on phantom 'jumper_runs' table rolls back the whole command

engine/cli/workflow_cmd/lifecycle.py:96 (cmd_terminate) and :310 (cmd_reset) run UPDATE jumper_runs SET status=... WHERE id=... — no 'jumper_runs' table exists in ANY migration (real tables are jumper.r_<queue> with columns run_id/state). Every invocation with a jumper_run_id raises UndefinedTable; the generic except catches it AFTER the workflow_run UPDATE but BEFORE conn.commit(), so the whole transaction rolls back — the commands do NOTHING and print an error. The ops emergency-stop path has silently never worked. Fix: replace with jumper client calls (cancel_task / fail-run path) or correct table+column names per queue; add a smoke test that exercises hwe terminate end-to-end (this class — raw SQL drifting from schema — evaded all tests because CLI paths have none).