Implement ConvexHull op for binary images

Currently, we only have two `geom.convexHull` ops:

Mesh out, double epsilon <= ConvexHull(Mesh in)
Polygon2D out <= ConvexHull(Polygon2D in)

and they're not too easy to use when you just have a binary mask image.

Let's create a new one that takes an `Iterable`:

RandomAccessibleInterval out <= ConvexHull(Iterable in)

See the [related discussion on gitter](https://gitter.im/imagej/imagej-ops?at=5b3f4e353572e970c17b001e).

Start with the existing geom.convexHull operations for Mesh and Polygon2D, then review the related Gitter discussion for the binary-image use case. The work is complete when a ConvexHull operation accepts Iterable input and returns a RandomAccessibleInterval output for binary masks.

Written by the indexing model from the issue text.