Merge the open CUDA PRs and release feectools on cunumpy 0.6 #686

Merge the open CUDA PRs in stack order, together with the feectools update, and release.

• CUDA PR 17 - struphy on the feectools CUDA stack #680 (PR 17: feectools 0.3.1 with the CUDA stack)
• CUDA PR 18 - Geometry evaluation on the GPU for all analytic mappings #681 (PR 18: geometry evaluation for all analytic mappings)
• CUDA PR 19 - Spline mappings on the GPU #683 (PR 19: spline mappings)
• Use cunumpy 0.6.1 #685 (PR 20: cunumpy 0.6.0) together with Update to cunumpy 0.6.1 feectools#94 (feectools on cunumpy 0.6)
• Release feectools with cunumpy 0.6, and point struphy's feectools submodule and the pyproject.toml pin at that release.

Part of #650 (see CUDA_STRATEGY.md).