allow varying size inputs, with varying size channels

[fdan]

This change addresses a todo item outlined in MLClient.cpp:

// Parse image. TODO: Check for multiple inputs, different channel size

The previous code enforced a single format on all inputs. I needed to make this change in my own fork and thought I'd submit it to Foundry for comment.

[ringdk]

Hello! Thanks for submitting the merge request. I can see what you're looking to achieve, however there are a couple of things that need to be addressed:

• The reason for using the format() and not the inputIop's bounds is because the format gives us the full size frame, and not just a data window. As you've found, it's not ideal.
• When changing it to imageBounds you need to be careful about handling negative values, and computing the correct width & height for the buffer. (eg. on your lines 467 & 482, it is possible for the values for fr and ft to be negative)
• Similarly, the buffers could be allocating more memory than is needed, i.e. you probably want [fr-fx-1, ft-fy-1] for the dimensions
• The data window isn't being passed through to the server
• The renderOutputBuffer() function is still parsing using the returned image using the format and not the imageBounds

Flagging this to @johannabar @guillaumegalesfoundry , as I'm wondering if ImagePlane can take more than 4 channels.

Thanks again @fdan !