Started to work on extracting label multisets from BigCAT

[hanslovsky]

Hi @shrucis1,
we finally started to work on extracting the label multisets from BigCAT (long overdue). This issue is just a FYI for you so you know what we are doing.

These are the relevant repositories:

• label-multisets
• n5-label-multisets

I also forked your

• bigcat-spark-downsampler, and
• hdf-n5-converter

I modified your code to use the new dependencies (label-multisets, n5-label-multisets) and to adapt to changes in other dependencies, e.g. n5. I also replaced jcommander by picocli, a more modern command line argument parser for Java (so far I am really convinced by it) and changed the high level structure: Now, bdv.bigcat.spark.SparkDownsampler creates a whole scale pyramid (s0 ... sN) at once (as defined by parameters). In n5, this will look like this:

+- /group/dataset/
|  s0/
|  s1/
...
|  sN/

We do not have an input dataset anymore but s0 is required to be present (can be a symlink which is also a parameter now). s0 is LabelMutlisetType already (and not any kind of Integertype), which can be created through your hdf-n5-converter (I modified that as well, see link above). I will sparkify the hdf-n5-converter as well, so it can be used for larger datasets.

Eventually, we will rename this repository, because it is not specific to/does not depend on BigCAT (anymore). Something like label-multisets-downsampler or label-multisets-downsampler-spark would probably be a good choice.