Allow normalization on both sides

[melix]

Expected Behavior

It should be possible to have expected output normalization, just like we have actual output normalization.

Current Behavior

Currently, only the actual output is normalized.

Context

In dependency management, we often have quite large outputs of dependency graphs, and often from one version to another, we have insignificant chances like number of spaces, due to the fact that this attribute has been renamed or whatever. So often tests are failing because this number of spaces in the expected output has changed (or additional blank lines, or whatever). It would be nice if we could just apply normalization on both sides, so that when we "search and replace" in a sample, we don't have to bother aligning text ourselves.

The idea would be that if norm(output)==norm(expected output), then we're good, and not only if norm(output)==expected manually normalized output.