Hillas calculation failed for some telescopes in camera-frame

[TjarkMiener]

Describe the bug
I'm using the latest ctapipe version (v0.16.0) with a subsample of Prod5b. When running the process tool, the Hillas calculation failed (either zeros or nan) for some telescopes in camera-frame.

To Reproduce
Steps to reproduce the behavior:

1. $ ctapipe-process -i /data3/users/nieto/Prod5b_sample/Prod5b_LaPalma_AdvancedBaseline_NSB1x_gamma_North_20deg_DL0_1e6evt/gamma_20deg_0deg_run126___cta-prod5b-lapalma_desert-2158m-LaPalma-dark.simtel.zst --progress --write-images --write-parameters --write-showers --camera-frame

Expected behavior
Hillas calculation should be also working when camera-frame is selected.

Supporting information
Screenshot with corrupted tables for tel 3 (LST) and tel 6 (MST with NectarCam):

Provenance log:
ctapipe-process.provenance.log

Additional context
It would be also great to store the information about the camera-frame in the meta data and just keep one naming convention for the parameter tables, i.e. removing the prefix "camera_frame_"

[maxnoe]

I suspect the problem is that for these telescopes the first event is not reconstructible, we then provide the wrong default container and that messes up all following events.

Let me check

[maxnoe]

Yes, this is the case. We return the global DEFAULT_IMAGE_PARAMETERS, which always contains the telescope_frame variant.

I'd solve this by just going ahead and removing the --camera-frame option to be honest and not add further complication for something we do not intend to support anymore.

[maxnoe]

On the other hand, the fix wasn't too hard...