proposal: add optional channel_names argument to regionprops

[VolkerH]

Description

I noticed that as of #5037 regionprops can measure intensity properties for multichannel images.
The resulting intensity properties are named with an appended -0 ,.... -n for a n+1 channel
image, e.g. mean_intensity-0, mean_intensity-1.

In practical applications the channel may have meaningful names, e.g. DAPI or GFP for fluorescence microscopy images.
My proposal is to add an optional argument channel_names to regionprops and regionprops_table.
If channel_names is None the behaviour should be unchanged, i.e. add an integer channel number.
If channel names is a Tuple or List of [str] of the same length as the number of channels, e.g. ("DAPI", "GFP") the appendices to the property name should be taken from that list, that is mean_intensity-DAPI, mean_intensity-GFP in the given example.

This would potentially have side effects for #5213

[VolkerH]

Oops ... after following the link to #5037 I realize that this isn't that trivia as I thoughtl because the numbering comes from the way that vector-valued features in regionprops are unrolled into regionprops_table.

[grlee77]

Thanks @VolkerH. I haven't inspected the implementation in terms of what refactoring may be required, but the idea seems reasonable to me.

[VolkerH]

looking into this a bit more, the new optional channel_names in regionprops and regionprops_table could be passed to _props_to_dict which would be the function in which the new functionality is added. A channel_names are then the default None would also flag that these are multidimensional intensity measurements.