`read_kilosort_as_analyzer` : read original `spike_templates.npy` instead of `spike_clusters.npy` from kilosort output

[pas-calc]

We should read the spike_templates.npy file in read_phy

Alternatively I propose to allow an argument to BasePhyKilosortSortingExtractor init that tells whether to try to use spike_clusters.npy if found or not and add an argument which was used.

The terminology should not be misleading because kilosort output can be manually refined by phy

spikeinterface/src/spikeinterface/extractors/phykilosortextractors.py

Line 341 in 25940b4

sorting = read_phy(phy_path)

#4366 (comment)