Jungfrau: current status

Jungfrau scans are currently being triggered through the web-gui (https://i24-daq-ui.diamond.ac.uk/). BlueAPI is hosted on the beamline cluster and is using numtracker integration and identity-backed authentication. See [here](https://gitlab.diamond.ac.uk/controls/containers/beamline/i24-services/-/tree/main/services/i24-blueapi) for blueapi configuration.

The live code is currently in `/dls_sw/i24/software/bluesky/scratch_k8s` . (There are currently some uncomitted hotfixes so be careful in that directory!).

### Next steps
1. There are some known issues which have been hotfixed in the live code but need proper fixes merged in. See [here](https://github.com/DiamondLightSource/mx-bluesky/issues/1643) for the full notes. In brief, we need to: change the parameter model so that `storage_directory` isn't being used when numtracker is being used; make device names in device composites consistent with i24's device factory names; investigate an IOC race condition where the IOC is turning pedestal mode off a short time after bluesky code sets pedestal frames and loops signals*; change `gui_run_jf_rotation_scan` param names and return type so that it is triggered correctly via web gui; put the jungfrau det Y position in a config file and in bluesky access this using the config server
2. Get ispyb integration working for rotation scans. There's an untested PR for this [here](https://github.com/DiamondLightSource/mx-bluesky/pull/1638)
3. When FastCS Odin is ready, we should use this instead. We can use much simpler code for this, instead of the `CommissioningJungfrau` device we can just use the `Jungfrau` device in ophyd-async, and no longer be using our temporary filewriter class
4. During next shutdown, migrate from numtracker-staging to numtracker. Can ask Peter H for help with this, it should be quite straightforward.
5. Get fixed target serial plans working. I have hacked together something [here](https://github.com/DiamondLightSource/mx-bluesky/tree/jf_ssx_fixed_feb_26) to show how to get started. I think this also needs some work on the web-ui end.

*In the meantime, if pedestal mode is being used, we need a 0.3s sleep immediately after detector is prepared, followed by another set to pedestal mode.

### Recurring issues
- Temporary filewriter IOC has had a few issues when talking to ophyd-async. This can cause strange symptoms in bluesky. If there are any problems appearing with the filewriter PVs then it's best to look at the IOC debug logs with Nick
- JF IOC breaks if we try to adjust detector settings after pedestal mode has been switched on, so we must make sure this is turned on as the very last thing, and that we switch it off at the end. In Bluesky code this seems to be stable, but there seems to be a race condition in the IOC (see above) which is still causing issues.