[ENH] docker containers for large scale maintenance and experiment tasks

[fkiraly]

Issue listing useful docker containers for large scale maintenance and experiment tasks:

• running check_estimator on all estimators or a subset thereof (e.g., specified by scitype). Runtime performance and memory benchmark, saves results by estimator, test and/or test-fixture combination as returned by check_estimator. With multiple reruns.
• running evaluate or benchmark experiments on a selected subset of estimators, producing a typical research paper benchmarking table. We should start with the forecasting benchmark class, as other tasks will move to a similar design.

FYI @benHeid, @hazrulakmal due to benchmarking; @astrogilda due to research; @shlok191 due to recent discussion; @YelenaYY due to project, and @SaiRevanth25 due to https://github.com/sktime/sktime/issues/6555

[shlok191]

@fkiraly, I'd love to work on this!