Correct configuration dataset creation

[stan-dot]

Scientists on the beamline know when the shape of the graph is right as expected, and that then the beamline configuration is correct.

It is said to be a long-developed high level skill that our control systems cannot replicate at the moment.

The shape of a problem looks like for a Reinforcement Learning use case. This would need to start with data procurement - necessarily by recording those human decisions. Some button and capture pipeline could record mappings of desired config - data quality - human evaluation.

[callumforrester]

@stan-dot Presume you want to write an adaptive plan to do this work? If so, is this not an issue for https://github.com/DiamondLightSource/i18-bluesky/?

[stan-dot]

I guess, though I had i22 in mind