Adding optima to tabular benchmarks (LCBench)

[Neeratyoy]

For tabular benchmarks, this is a must-have option.

Can also be added as a property to the Result object.

Currently, I want to access this as a benchmark property ideally and not have to query for this. Hence, adding it for each row.

Happy to discuss the best way to retrieve it.

This option can be kept general and continuous benchmarks can have this as None. Also could be used as a proxy check to test if a benchmark is tabular or surrogate.

[eddiebergman]

Going to close this since it's superseed by the new metrics API, however I will add this in!

[eddiebergman]

The new method will be:

# Only available on tabular datasets, uses entries in the table
benchmark.table_optimums["bal_accuracy"].error # for minimize optimum
benchmark.table_optimums["bal_accuracy"].score # for maximize optimum

# Available on all datasets, just takes the metrics definition
benchmark.metric_optimums["bal_accuracy"].error # for minimize optimum
benchmark.metric_optimums["bal_accuracy"].score # for maximize optimum