Add Python bindings - shared memory

[icui]

Create a Python package that allows communication between Python and SPECFEM code. The usage include

• Monitor execution status without writing files
• Read and visualize wavefield while SPECFEM is running
• Control (e.g. pause, resume, terminate) the specfem execution
• Save wavefields and seismograms as custom format
• A dedicated specfem folder is no longer necessary, which is easier for workflow managers

Two possible implementations:

1. Use shared memory. Add a runtime parameter that instructs SPECFEM to use the assigned memory address instead of allocating new CPU memory for the wavefield. This requires minimal change to the current SPECFEM but is less flexible on the Python side.
2. Expose a function that takes HostView as input, e.g. run_specfem(HostView wavefields). From main(), we call run_specfem() with default initialized host memory whereas from Python code, we call run_specfem() with Python-initialized host memory.

[icui]

It seems that pybind is a more standard approach for kokkos views https://github.com/kokkos/pykokkos-base?tab=readme-ov-file https://www.youtube.com/watch?v=1J3JQ3d3cRc&feature=youtu.be&t=2115

[icui]

Also, it's worth thinking how this should be implemented with MPI

[icui]

Package created in PyPI. https://pypi.org/project/specfempp/