Better support for N-dimensional data

[crueden-x]

Eliminate IFormatReader.getChannelDimTypes and IFormatReader.getChannelDimLengths, in favor of more comprehensive getDimTypes and getDimLengths methods, which return all used dimensions in correct rasterization order.

• OME dimensions include: X, Y, Z, C, T
• Examples of non-OME dimensions:
  • lifetime bins
  • excitation wavelength
  • polarization
  • screen
  • plate
  • well
  • field

Advantages:

• fully capable of expressing N-dimensional image data
• allows for XZ scans and other spatial configurations
• eliminates the need for DimensionOrder? and Interleaved core metadata

Work to do:

1. Rework ChannelSeparator? a little
   • It should always "pull" XY to the front

2. Rework DimensionSwapper?; it is more complex now

3. Rework DataBrowser? to use new methods, displaying extra sliders as before

4. Rework FileStitcher? and AxisGuesser? a bit
   • Guessed axes become new dimensions

5. Change anywhere else that calls getChannelDim* methods

6. InCell? reader can be retrofitted to use this new functionality
   • Certain other readers can benefit as well: Flex, Fluoview, ...

7. Initially, conversion to OME-TIFF with bfconvert should throw an exception if any dimensions beyond 5D exist, since there is nowhere in the OME-XML schema to store the extra dimensional information

[jmoore]

imported from ​bio-formats:#398