removing autoinsertion of rescale_to_int for sweep runs

[dkazanc]

Fixes #520

As it is stated in the issue #520 the auto-insertion of rescale_to_int in sweep runs where Paganin and reconstruction involved makes the reconstruction invalid as the input data for FBP is converted to integer type. This has been resolved by removing auto-insertion of rescale_to_int but enabling a conversion of floating point data in the image saver as oppose to just break previously when the data type is not uint8/16/32. This change shouldn't have any effect on the normal runner when the auto-insertion of scaling is not performed. However, when the user doesn't scale the data before the image saver the run will not fail, but the data won't be scaled uniformly.

It is worth to note that when the users would like to save the rescaled projections themselves into images, the data will be rescaled into uint making processing after invalid. So, essentially we might want to restrict the number of rescale_to_int usages present in the pipeline (1 is the limit).

Checklist

• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• I have made corresponding changes to the documentation

[dkazanc]

the errors in tests should go away once httomolib library is updated with this PR

[dkazanc]

iris is down apparently, the tests are passing for me locally and httomolib version has been bumped to v.2.2 with the relevant changes so I will merge this.