testSIRF.TestGradientMR_2D failure. increase tolerance?

[KrisThielemans]

FAIL: test_TVdenoisingMR (test_SIRF.TestGradientMR_2D)

https://github.com/SyneRBI/SIRF-SuperBuild/actions/runs/6956888226/job/18931696600?pr=846#step:8:155

Mismatched elements: 74722 / 131072 (57%)
Max absolute difference: 0.14627695
Max relative difference: 0.3511309
 x: array([[[0.198+0.j, 0.195+0.j, 0.191+0.j, ..., 0.203+0.j, 0.202+0.j,
         0.199+0.j],
        [0.198+0.j, 0.196+0.j, 0.193+0.j, ..., 0.204+0.j, 0.202+0.j,...
 y: array([[[0.196+0.j, 0.195+0.j, 0.193+0.j, ..., 0.204+0.j, 0.202+0.j,
         0.201+0.j],
        [0.196+0.j, 0.195+0.j, 0.194+0.j, ..., 0.203+0.j, 0.202+0.j,...

[paskino]

This is due to the change in TotalVariation where the default setting is to use warmstart=True. https://github.com/TomographicImaging/CIL/blob/0418ed2abd42c085db25aa3d0405b4b16e6591e4/Wrappers/Python/test/test_SIRF.py#L254C53-L258

In this case the second call does not start with an uninitialised TotalVariation.