114 tests are failing on gpu machine

[jank324]

Description

This PR overhauls the way devices and types are handles by Elements and Beams. The idea is to make it more in line with how original PyTorch Modules like nn.Linear do it (https://pytorch.org/docs/stable/_modules/torch/nn/modules/linear.html#Linear). This also addresses the fact that according to (https://stackoverflow.com/questions/58926054/how-to-get-the-device-type-of-a-pytorch-module-conveniently) Modules shouldn't have overarching type and device properties because their parameters and buffers may be on different devices and of different types.

On a more practical note, this makes the entire device system much more robust and basically completes MPS support from Cheetah's side (#61 / not all operations we use are implemented in PyTorch for MPS yet).

A minor downside is that we no longer do automatic device selection. This does, however, make that part of Cheetah more predictable (I previously had a lot of code using Cheetah fix the device because weird things would happen).

This PR also prepares an eventual fix for #113.

Motivation and Context

• I have raised an issue to propose this change (required for new features and bug fixes)

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation (update in the documentation)

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).

Note: We are using a maximum length of 88 characters per line

[cr-xu]

If I read the changes correctly, now we can safely switch between [float16, float32, float64] for elements, right?

[jank324]

I haven't tested it (other than quickly trying to track through a quadrupole with MPS, see #61 ), but in theory, we should now be able to safely switch between all dtypes and devices, yes.

[cr-xu]

I think it's worth mentioning explicitly in the documentation that it actually changes the behaviour of element creation, i.e. one can now create elements conveniently as before v.6.0, without the requirement to wrap every parameter as tensor first

dipole = cheetah.Dipole(length=0.1, angle=0.1, dtype=torch.float32)

[jank324]

I think you are right, but I also think this should be tested before making that claim.