feat: add basic sp selection score modelling

[rvagg]

Getting this work up in a Draft for now so I can pause and think about something else for a little while.

This is an attempt to create a simulation of a set of retrievals in such a way that we can tweak the scoring mechanism (weights etc.) to see how/if they make a difference.

This works, (run main in pkg/session/model/cmd), but the averaging of provider behaviour across runs is causing any tweaking to get lost in the wash. I need to set up a static list of providers with fixed(ish?) characteristics up front so they can be sorted in more meaningful ways and their score comparisons have more impact.

[codecov-commenter]

Codecov Report

Merging #298 (1bcd218) into main (fe9e760) will decrease coverage by 3.29%.
The diff coverage is 3.28%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #298      +/-   ##
==========================================
- Coverage   72.55%   69.27%   -3.29%     
==========================================
  Files          68       72       +4     
  Lines        6217     6519     +302     
==========================================
+ Hits         4511     4516       +5     
- Misses       1458     1753     +295     
- Partials      248      250       +2     

Impacted Files	Coverage Δ
pkg/session/model/cmd/main.go	0.00% <0.00%> (ø)
pkg/session/model/pop.go	0.00% <0.00%> (ø)
pkg/session/model/prob.go	0.00% <0.00%> (ø)
pkg/session/model/sim.go	0.00% <0.00%> (ø)
pkg/session/state.go	92.11% <90.90%> (+0.30%)	⬆️

... and 2 files with indirect coverage changes