Integrating custom NBLAST scoring matrices

[schlegelp]

We have recently generated some custom NBLAST scoring matrices which turned out to be very useful and I'm thinking about how to best package them up for easy use by others. I wouldn't want to put them into navis directly but e.g. flybrains might be a good place to put them.

This would already work:

import flybrains
scores = navis.nblast_allbyall(neurons, smat=flybrains.my_matrix) 

To make things easier we could additionally allow a syntax like this:

scores = navis.nblast_allbyall(neurons, smat="flybrains:my_matrix")

where a string with a colon is interpreted as "get this object from this module" and users don't even have to import flybrains themselves.

Another option would be to implement a plugin system similar to what we're doing for transforms:

# flybrains/scoring.py
mat = pd.read_csv('../data/scoring_function.csv')

navis.nbl.register_scoring_matrix(mat, name="flybrains_custom_matrix", description="...")

import flybrains
scores = navis.nblast_allbyall(neurons, smat="flybrains_custom_matrix")

We could even use Python's entry_points to avoid having to explicitly import flybrains (see octarine3d's plugin system). The advantage of the latter would be discoverability: we could have a function to list all available scoring functions.