Implement Benchmarks: Mono geometry reco, and Stereo Geometry from mono-reco

This needs a trained mono reco model that has been applied to all events, so can't do much yet.

I think the metrics for both mono-reco for each telescope type and stereo reco from weighted mono-reco are identical to the metrics for #19 , it's just that the input columns are different. So the Benchmark can just re-use the same Metric classes, but fill them differently.