CUDA Support for Z4C Rhs kernels and grid solver (FMR and Fixed grids)

[at0m741]

Summary

This update merges the moving-puncture-regrid work into the current CUDA GPU support branch.

Highlights

• Added the new Z4c-oriented moving puncture workflow, replacing the older BSSN test layout with z4c
  test targets and runners.
• Added moving-puncture regridding support, including nested and BBH split fixed mesh refinement
  layouts.
• Added FMR memory pooling utilities to reduce allocation pressure during moving puncture evolutions.
• Added optional HDF5 export support for simulation and diagnostic output.
• Improved moving puncture boundary handling, including Sommerfeld/radiative boundary configuration and
  sponge support.
• Preserved and integrated the CUDA backend path for Z4c/BSSN RK stepping, CUDA RHS kernels, and CUDA-
  related regression tests.
• Added/updated Python bindings and Python-side Z4c helpers.
• Updated CMake targets and test organization around the new z4c naming.

Validation

Built successfully with:

• cmake --build build-codex-cuda -- -j8
• cmake --build build-codex-check -- -j8

Both CUDA and non-CUDA builds compile successfully.

Merger t = 15.3262

[at0m741]

merged