Add remove_stripe_fw with memory peak calculator

[mfep]

• Copied remove_stripe_fw implementation from https://gitlab.esrf.fr/tomotools/nabu/-/blob/master/nabu/thirdparty/tomocupy_remove_stripe.py#L311
  • Updated the original implementation so that trivial computations are not offloaded to the GPU
  • Note, that the implementation uses the PyWavelets dependency, which was already part of the httomo dependency chain, so this implementation keeps using it
• Added simple regression test to check that the result of the algorithm does not change
• Added simple regression test in zenodo-tests
• Implemented inline device memory peak calculation, which can be run by a bool flag passed to remove_stripe_fw
  • The memory peak calculation does not perform the stripe removal and does not use the GPU at all
  • The inline implementation's advantage is that the memory estimation path can reuse the same computed values that the regular path would use. Otherwise it would be difficult to estimate the memory peak of an iterative algorithm like this
  • A unit test is added for the calculator, that proves that it is accurate with a margin of +/- 1%