Jungfrau plans sometimes fail to apply and remove pedestal mode

A couple of issues I ran into running the WIP jungfrau collections. These were running on top of commit https://github.com/DiamondLightSource/mx-bluesky/commit/18d7c87221a2ece9e33187d4f7893da4f5f93875. I can't rule out the IOC being the cause of some of these, and unfortunately don't have IOC logs from this testing.

- Saw a case where when running pedestal collections, the detector wasn't put in pedestal mode. Example log; although it looks like this is setting the mode on 48ms after switching it off, the detector wasn't in the mode when it was actually triggered. [2025-10-16T113313_pedestals_dont_activate.log](https://github.com/user-attachments/files/22966548/2025-10-16T113313_pedestals_dont_activate.log). Reading this, it does look as though it is trying to turn it on...

- Sometimes, the pedestal mode doesn't get turned off when finishing a pedestal collection. [2025-10-16T113438_pedestals_remain_on.log](https://github.com/user-attachments/files/22966536/2025-10-16T113438_pedestals_remain_on.log), [2025-10-16T112734_pedestals_remain_on.log](https://github.com/user-attachments/files/22966663/2025-10-16T112734_pedestals_remain_on.log). Leaving the detector in this special mode probably wouldn't be such an issue, except that I needed to run manual darks as the IOC appeared to be setting short exposures wrong. Probably we don't want to leave it in this state?

With these latter cases, I don't think the detector actually ended up in an idle state; we needed to manually `sls_detector_put rx_stop` afterwards. I didn't get enough diagnostic time to narrow down the causes of this; possibly related to issues where the values aren't set (possibly silently?) if the detector isn't idle when the PV is attempted to be written.

## Acceptance Criteria
- Detector pedestal mode is always activated when running pedestal collections
- Detector is taken out of pedestal mode at the end of pedestal collectios