I noticed that during a lot of the ALS_DARM feed-back trials last night the suspension length control signal wasn't immediately shut down after lock-loss, leading to a 20sec period of buzzing the test masses. That can't be good for keeping energy out of the suspensions. I updated the guardian to immediately cut down the suspension drive signal upon lock-loss:

- isclib has a function killDrive(), which is called right after lock-loss to cut the drive
- It uses TRAMP=0 & GAIN=0, because that combination is faster than an ezca.switch command, which has to wait for a read-back (and can take a good fraction of 1sec to execute)
- Currently is kills the ETMY_MN_LOCK_L and the ETMY_TM_LOCK_L drives
- killDrive is called first thing in DOWN. that should catch must unexpected cases
- killDrive is also called directly from the lock-checker decorator in ALS_DARM decorator, which should be even faster.
- However a Guardian solution like this stops working when the guardian stops working or is paused. We might want to consider a fast VIS drive shutdown in the front-end code as well.