Use massively parallel sundials solvers for running many solves with different input parameters

[jsbrittain]

Description

Implement GPU-enabled solvers within Idaklu for massively parallel runs. Designed to provide support for input parameter sweeps, etc.

Pull request is currently work in progress:

• Refactor Idaklu code to support subclasses specifying separate solvers. Current solvers follow a consistent architecture, so all inherit from an OpenMP class. This permits flexibility and a cleaner extension for, e.g. GPU solvers.
• Implement a GPU solver (CUDA) within the framework.
• Debug GPU solver (!).
• Provide a pybamm/python interface to access new functionality.
• Ensure that GPU functionality is optional on build (i.e. Idaklu builds with or without CUDA support, as needed).

Fixes #2644

Type of change

Please add a line in the relevant section of CHANGELOG.md to document the change (include PR #) - note reverse order of PR #s. If necessary, also add to the list of breaking changes.

• New feature (non-breaking change which adds functionality)
• Optimization (back-end change that speeds up the code)
• Bug fix (non-breaking change which fixes an issue)

Key checklist:

• No style issues: $ pre-commit run (see CONTRIBUTING.md for how to set this up to run automatically when committing locally, in just two lines of code)
• All tests pass: $ python run-tests.py --all
• The documentation builds: $ python run-tests.py --doctest

You can run unit and doctests together at once, using $ python run-tests.py --quick.

Further checks:

• Code is commented, particularly in hard-to-understand areas
• Tests added that prove fix is effective or that feature works

[valentinsulzer]

@awadell1 @pghege this might be of interest (GPU solvers for battery models)

[jsbrittain]

Development enabling GPU support for the IDAKLU solver stalled since parallelisation required (gpu) device-kernels to be compiled from PyBaMM models. This would require compilation of casadi libraries onto the device for execution, or implementation / translation of an alternative expression tree model (with interface) to the solver. At this time the PR is being closed in favour of the approach adopted in #3121.