Systematise organisation of internal interaction and callbacks

[dperl-dls]

Currently, we have a situation where zocalo_interaction contains the interaction and the callback code, while write_nexus and ispyb_interaction are independent, and the ispyb and nexus callbacks are together in one file. this is horribly unsystematic.

I propose that we separate the zocalo callback back out from the zocalo interaction file, and split up the callbacks from each other, so that:

• all the code needed for zocalo interaction (for any type of experiment) is in zocalo_interaction
• all the code needed for ispyb interaction (for any type of experiment) is in ispyb_interaction
• all the code needed for nexus writing (for any type of experiment) is in write_nexus
• and the callbacks are all individual for experiment type, eg: gridscan_ispyb_callback, rotation_zocalo_callback etc.

That way it's the same for everything, we minimise duplication of code for the actual interactions, and minimise checking what the experiment is in the callback, just subscribe to the correct one in the plan.

Any thoughts?