feature sets table

[tibuch]

This PR adds so called ComputerSets (<==> FeatureSet). A ComputerSet holds different ComputerOps which build a common known measurement-group.

Now all FirstOrderStats ops could be combined in a FirstOrderStatsComputerSet. Same goes for different groups like Zernike, Histogram or Geometric2D (and many others which have to be implemented).

There are non-configurable and configurable ComputerSets. Configurable means that each op of this ComputerSet could either be active or inactive.

The ComputerSets are processed by the ComputerSetProcessor. There are three different implementations. One for ComputerSets based on Iterables (FirstOrderStats), one for ComputerSets based on LabelRegions (Geometric2D) and one for ComputerSets which should be computed over ROIs (FirstOrderStats on every segment of a LabelRegions).

To a ComputerSetProcessor arbitrarily many ComputerSets could be added. The ComputerSetProcessor will create a Table with a column for each computed op of every ComputerSet. The IterableComputerSetProcessor creates a table with one output row and the LabelRegions- and ROIComputerSetProcessor creates a output row for each segment of the Labeling.
The ComputerSetProcessor takes care of unique column names in the output-table.

// create a FirstOrderStatsComputerSet with all Computers activated
FirstOrderStatsComputerSet<Iterable, DoubleType> stats = ops.op(FirstOrderStatsComputerSet.class,
                Iterable.class);

// create a FirstOrderStatsComputerSet with only Mean and Max Computers activated
FirstOrderStatsComputerSet<Iterable, DoubleType> stats1 = ops.op(FirstOrderStatsComputerSet.class,
                Iterable.class, Arrays.asList(new Class[] { Mean.class, Max.class }));

// create an IterableComputerSetProcessor which holds the two ComputerSets
        IterableComputerSetProcessor<FloatType, DoubleType> processor = ops.op(IterableComputerSetProcessor.class,
                Iterable.class, new ComputerSet[] { stats, stats1 }, new DefaultComputerSetTableService<>());

// compute the ComputerSets on img
        Table<Column<DoubleType>, DoubleType> result = processor.compute1(img);

@dietzc and @ctrueden please take a look.

[dietzc]

this problem can be resolved as soon as we have introduced the new or better matching which support proxy objects.

[tibuch]

Should I rewrite the history again?

[dietzc]

I think thats fine. Let's wait with merging until after the hackathon, such that we potentially get rid of the Matcher workarounds?

[tibuch]

I rebased this branch. Maybe we can merge this PR soon :)

[ctrueden]

@tibuch This is still waiting on the generic type matching, right? I'll add a label assuming so.