[ENH] Logging intermediate results in forecasting benchmarking

[benHeid]

Is your feature request related to a problem? Please describe.
When applying long benchmarks with multiples of forecasters etc. It would be cool if intermediate results would be stored. This would have the following advantages:

• In case of failures, not all benchmarks need to be executed again.
• It would be possible to take a look into preliminary results and see how many benchmarks are already executed.

[fkiraly]

I think the first design by @ViktorKaz and myself, for classification, has that feature. It's fallen a bit out of step with the rest of the code base, but perhaps this might be a design inspiration.
(current benchmarking module)