Properly implement mask channels

Currently, the mask channels in ChannelViewMaskWidget are hard-coded to align with the image channels. In principle, they can be independent.

In this case, the only information ChannelViewMaskWidget would need about the runtime context of ChannelViewWidget is the image size and the axes to add the masks into. This leads to a beneficial decoupling, and the new mask widget can get a more generic name and is better composable.