[ENH] simple collation of summary results into an array

[TonyBagnall]

Describe the feature or idea you want to propose

My basic use case is

1. Run all 112 on cluster for algo X
2. Compare X performance to reference results

it would be helpful to have a simpler results collators from raw results to make it easier in (2) to compare and plot against reference results.

Describe your proposed solution

something that just extracts a dictionary of {problem: mean_accuracy} (or other measures) in a format compatible with get_estimator_results in aeo would be useful.

Other comments and alternatives considered

No response