Rework ROI detection

[Lyr3x]

The ROI detection needs to be reworked. It should find the best ROI settings defined by distance and target surface size.

[Lyr3x]

I need to test this thoroughly. I am not sure right now if this will improve things or make them worse. From my latest findings I think the ROI size itself does not matter as much as the target surface material and sensor alignment.

[Lyr3x]

I will remove the rewrite from the v1.3 milestone and will only fix the current implementation. This topic needs further research and in depth testing with various feedback loops.

[Lyr3x]

I implemented a fairly simple change to produce one of the following ROIs

• 4x8
• 6x12
• 8x16

I am not sure right now if that helps in all scenarios and I need to do more testing. Feedback is much appreciated!

[Lyr3x]

ROI center settings must be reworked too as there are three different situations where the ROI center points need to be set. Currently the center is not working properly everywhere

[CarsonF]

New schema suggestion

cv.Optional("roi", default={"height": 16, "width": 6}): cv.Any(
    cv.one_of("auto"),  # formerly "roi calibration"
    cv.Schema({
        cv.Optional("height", default=16): cv.int_range(min=4, max=16),
        cv.Optional("width", default=6): cv.int_range(min=4, max=16),
    })
),

Maybe default switches to auto once the implementation is deemed better