Fix failing tests

[xEnVrE]

Some of the tests are failing due to small mismatches between the ground truth rendering and the one generated by the test.

We should investigate whether this is due to the non repeatability of the rendering tasks.

[claudiofantacci]

How much is the mismatch error?

[xEnVrE]

It depends. For example, test_scissors reports a mismatch of only 1 unit for the green channel

[Utils][compareImages] Found non-equal pixel.
[Utils][compareImages] Pixel at (12,65) is [255, 128, 51].
[Utils][compareImages] Should be [255, 127, 51].

Adding a threshold in the compareImages util does not help since, after that, other mismatches were popping up in the order of 10 units.

By looking at the ground truth and generated files, it seems that there is a shift, that is noticeable if one alternates rapidly the figures using an image viewer, in the pixels.

ground truth

generated

However, they seem pretty much the same to me!

[claudiofantacci]

The ground truth seems blurred.
I would try to generate the above image 10 times and see if the pipeline produces different results.
I would argue that and error of 1 could be tolerated (maybe some data conversion error?), but 10 seems way too off.