Refactor multi-channel image support

[hinerm]

The ImgToMatVector and MatVectorToImg convertors do not follow the Mat <--> Img converter format of having static methods we can call into for conversion. This limits them in being exposed e.g. via SciJava Ops.

We should:

• extract the conversion logic to static utility functions
• Reassess the conversion logic. This looks extremely easy to break, e.g. interleaved data. Also may break down with 4+ dimensions.

[ctrueden]

Relatedly, I was wondering whether we should have a separate imglib2-opencv component with just the static method converters, and make this imagej-opencv component only the SciJava Converter plugins on top... that way we can avoid imposing an ImageJ2-layer dependency on scijava-ops-opencv.