mlos_bench: first class scheduler support

[bpkroth]

Right now we lack support for complex scheduling policies of future trials in mlos_bench.

This can be useful for:

• repeating a trial across space (nodes, regions, etc.) and time in order to gain confidence in results and test for robustness to noise
• manually "focusing" on a region of the config space
• etc.

[bpkroth]

Wrote some notes for this in https://github.com/microsoft/MLOS/issues/380#issuecomment-1832711391

[bpkroth]

#676 adds some initial support for this