#436 BUG: durable_cron never fires - last_run variable not persisted before first sleep

## Problem The durable_cron tool never actually spawns child workflows. It keeps waking up, recalculating next_run, and going back to sleep. ## Root Cause In engine/tools/durable_cron.py: 1. last_run_time is retrieved with ctx.get_variable(..., default=NOW) 2. If variable doesn't exist (first run), defaults to NOW 3. next_run = croniter(cron_expression, last_run_time).get_next() = next minute 4. Since now < next_run, sleep_duration > 0, workflow sleeps 5. On wake, the variable STILL doesn't exist (only saved after spawn) 6. Defaults to NOW again, repeats forever ## Expected Behavior - calc_next checkpoint should return cached next_run time on wake - OR last_run should be saved BEFORE first sleep ## Actual Behavior - calc_next step re-executes each wake (checkpoint not honored) - last_run never saved (spawn never reached) - Workflow loops forever in sleep-wake cycle ## Steps to Reproduce 1. Submit a durable_cron workflow 2. Check logs - see 'Sleeping for Xs' messages every minute 3. Never see 'Time to run! Spawning target task...' message ## Affected Workflows - farshid_health_scheduler (57d1cff8-4407-4a8e-8e1a-6acb6a44c64b) - Any durable_cron workflow ## Additional Issues - No customer visibility (task_count=0, cryptic step names) - workflow_schedules.last_run_at never updated