Tidy up OAV devices and grid plan

[d-perl]

Some OAV device namings are confusing, e.g.:

• MJPG is called "snapshot"
• inconsistent use of input_plugin_pv and input_pv (maybe areadetector plugins should be a class...)
• chain plugins generically or at least in a consistent way
• structure of snapshot/snapshotwithgrid etc needs cleaning
• top_left_x_signal etc. in snapshot device classes - add clarity that this refers to grid and remove redundant "signal"
• take arguments for grid size, margin, # of boxes (eventually to come from json)
• deal with grid with no pin at zoom 0
• maybe average edge detection for stability, also maybe make sure rounding is correct so grid always extends to the edge of the sample
• add backlight manipulation
• oav_centring_plan should be in experiment_plans rather than device_setup_plans

also:

• look at bluesky visualisation options for live debugging

[DominicOram]

To address the above that have not been addressed in the PR:

• Tidy up MJPG and SnapshotWithGrid structure dodal#50
• deal with grid with no pin at zoom 0 - this seems like a way off, and will probably fall out of Full centring algorithm #499
• maybe average edge detection for stability, also maybe make sure rounding is correct so grid always extends to the edge of the sample - after tests on the beamline I think this was not much of an issue, we should get this on the beamline working and open a new issue if stability/accuracy become issues
• look at bluesky visualisation options for live debugging - can't remember what we exactly wanted to do with this, can pick it up if it's useful later