Space charge tests: improve backward-mode test ; add forward-mode test

[RemiLehe]

Description

This PR improves the space-charge tests:

• It now tests the accuracy of the gradient calculation, for the backward-mode differentiation.
• In addition, it adds a similar test for forward-mode differentiation.

Motivation and Context

Since Cheetah is built with pytorch, it supports backward-differentiation, but also forward differentiation (see this page). This could be useful since the scaling of forward differentiation is different from that of backward differentiation.

In addition, previous tests of backward-differentiation mode were simply checking that the differentiation ran without error, but they did not check the accuracy of the computed gradient. In this PR, we now test the accuracy of the gradient, by reusing the parameters of https://accelconf.web.cern.ch/hb2023/papers/thbp44.pdf, and by using an analytical formula to find the expected gradient.

Types of changes

New test

Checklist

• I have updated the changelog accordingly (required).
• My change requires a change to the documentation.
• I have updated the tests accordingly (required for a bug fix or a new feature).
• I have updated the documentation accordingly.
• I have reformatted the code and checked that formatting passes (required).
• I have have fixed all issues found by flake8 (required).
• I have ensured that all pytest tests pass (required).
• I have run pytest on a machine with a CUDA GPU and made sure all tests pass (required).
• I have checked that the documentation builds (required).

[RemiLehe]

This is reusing the same code as in the test test_cold_uniform_beam_expansion above.

[cr-xu]

Thanks for providing a forward AD test. I think similar tests could also be added to other parts, but we can do that in a separate PR.