Impact-distance is stored for all telescopes

[LukasNickel]

Describe the bug
Impact reconstructor stores distance for all telescopes in a subarray independent of if they triggered or not.
This can lead to situations where the dl1 table of a telescope is empty, but the dl2 table contains (useless) impact distances, which leads to confusing behaviour of the tableloader (Columns missing from the table, but no error).

Expected behavior
I expect only distances to triggered telescopes to be written.

Supporting information

ctapipe/ctapipe/reco/shower_processor.py

Lines 81 to 89 in 27e5439

	impact_distances = shower_impact_distance(
	shower_geom=event.dl2.stereo.geometry[k], subarray=self.subarray
	)
	for tel_index, impact_distance in enumerate(impact_distances):
	tel_id = self.subarray.tel_ids[tel_index]
	event.dl2.tel[tel_id].impact[k] = TelescopeImpactParameterContainer(
	distance=impact_distance
	)

Additional context
It mostly matters for small files, but our dl2 geometry file is such a candidate:

[maxnoe]

I just also ran into this. I agree, we should only store the triggered telescpoes, as this is an invariance assumed by the hdf5 table reader.