Replace run_somd.sh with Pydantic configuration class

[fjclark]

Description

This completes pretty much all the changes suggested in this comment: #11 (comment). All tests pass, although a3fe/tests/test_sys_prep_configuration.py::test_config_dump_and_load is flaky - it seems to fail when run as part of the full test suite. We should fix this before merging into main.

@Roy-Haolin-Du please could you take a look? Feel free to merge in if you're happy and fix any mistakes I've made. Pushing quickly to avoid any duplication of effort with you also working on it. Thanks!

[Roy-Haolin-Du]

A brief explanation to avoid misunderstanding.
Very helpful method involved "analysis_slurm_config"
When 'analysis_slurm_config' is passed as an argument from process_grads.py, it is assigned to renamed as 'slurm_config' to do analyse in mbar.py.

[fjclark]

Yes - the idea is store two SlurmConfig objects in all SimluationRunner objects. When running MD, the slurm_config attribute is used, and when doing analysis the SlurmConfig saved in the analysis_slurm_config attribute is used. Anything which runs MBAR is passed the analysis_slurm_config, which is the only slurm_config it knows about. There's a section in the docs which shows how it's used:

The molecular dynamics simulations should be run on GPUs - they are unbearably slow on CPU. However, you may want to run the MBAR analysis on CPUs to minimise submissions to the CPU queue. 
To do this, you can supply an ``analysis_slurm_config`` which is different to the ``slurm_config``, which will only be used for the MBAR analysis.

.. code-block:: python
    analysis_slurm_config = a3.SlurmConfig(
        partition="my-cluster-cpu-partition",
        time="00:10:00",
    )
    calc = a3.Calculation(
        slurm_config=slurm_config,
        analysis_slurm_config=analysis_slurm_config,
    )
  Then make sure to set ``slurm=True`` when running the analysis, e.g. ``calc.analyse(slurm=True)``.

Note that the slurm_config and analysis_slurm_config attrs of all sub-simulation runners just point to the corresponding attr of the top simulation runner (e.g. a calculation if you're just running one calculation), so you can change all the slurm options for all sub-simulation runners by doing e.g. calc.slurm_config.partition = "other_partition".

Hope that makes sense.