Issue/1070 second iteration changes for parallelproj interface

[markus-jehl]

Added a call to get the max_FOV_radius of the scanner, to fix the parallelproj interface for non-cylindrical scanners.

[markus-jehl]

Kris (https://github.com/UCL/STIR/pull/1074): "I suspect it isn't the image radius though. Could you check what parallelproj does if you give it a small radius? Instead I think we should check the return value of get_intersections_with_cylinder such that we don't ever use intersections with a cylinder where there aren't any,"

[markus-jehl]

I tested it, and parallelproj works as expected for smaller radii. I now also added handling of the return value of get_intersections_with_cylinder. The remaining things to be done are:

1. Add parallelproj sections to existing tests and confirm they pass.
2. Run Positrigo reconstruction pipeline to check there are no obvious bugs anymore.

[KrisThielemans]

Do you intend to add the tests in this PR? Maybe best to finish this one and squash-merge it first.

[markus-jehl]

Initially, I was thinking of doing it as part of this PR - but I agree it would be nice to get this merged in sooner rather than later!

[markus-jehl]

IMPORTANT: currently, parallelproj truncates the resulting images to a cylinder with the inner ring radius of the scanner. This means any activity in the corners of a BlocksOnCylindrical scanner will be removed.

[KrisThielemans]

AppVeyor time-outs