Feature Request: Making order = np.argsort(device_channel_indices) optional

Dear Spikeinterface community,
I am asking for this feature in the _set_probes() as I would like to feed my previous dataset sorted by kilosort standalone into an analyser with the new function read_kilosort_as_analyzer #4366
The problem is that my probe map, coming from probe_interface library is not sorted by channels. While using spikeinterface to do spike sorting, it is fine to sort them by channels. However, when using read_kilosort_as_analyzer, it will check whether channels location used in the kilosort4 phy folder matches with its counterpart in the Recording (e.g. OpenEphysBinaryRecordingExtractor). Since kilosort4 does not sort the map, there is always a mismatch between the two.

or

I can certainly create a set of probe maps and sort them by channels first before using kilosort standalone to do spike sorting so that I will never to worry about this mismatch. However, would it be possible to make order = np.argsort(device_channel_indices) an option to skip so that both Recording and either my original probe.prb or kilosort's channel_map.npy, channel_locations.npy, and channel_shanks.npy are equally unsorted?

Or if you have any suggestion about circuvment the mismatch in channels location when using read_kilosort_as_analyzer?

Many thanks in advance.

Start at _set_probes() and follow how device_channel_indices is ordered, then inspect read_kilosort_as_analyzer and the handling of channel_map.npy, channel_locations.npy, and channel_shanks.npy. Done means an unsorted probe and Kilosort dataset can be read without a channel-location mismatch while existing sorted behavior remains intact.