[discussion] feature matching of ebrains dataset should be reworked

[xgui3783]

Currently, feature matching of ebrains dataset is done via fuzzy match of region names and species (anchor in siibra compared to study target in KG). However, since each region has a uuid, the matching should be done with uuid matching. This has several advantages:

• non ambiguous parsing of region in anatomicalanchor. Currently, name and species is provided to anatommical anchor. The anatomical anchor needs to parse the region name (as a string), which is computationally intensive.
• we get, for free, semantic links, already established in the KG (homologies across species, prior versions, etc)
• currently region instances in siibra lacks id. This would introduce persistence reference to region instances, and brings it inline with the uuid of parcellation entity version (KG equivalent of region instances)

Potential blocking issues

• region instance in siibra does not map 1-1 to parcellation entity version in KG. hoc1 left for example, maps to 3 instances, one in each spaces