Kripke status

## Components
1. [x] Source code, Build system: CMake. Fixing race condition
2. [x] Source code released
3. [x] Spack package.py
4. [x] Ramble application.py
5. [x] Benchpark experiment.py
6. [x] +cuda experiment.py
7. [x] +rocm experiment.py
8. [ ] Benchmark dependencies: RAJA, Umpire, Chai (consider removing)
9. [x] small test size problem for CI
10. [x] Scaling (Weak, Strong, or both)
11. [ ] Large scale problem size defined
12. [x] CI running on CPUs/CDash
13. [x] CI running on NVIDIA GPUs/CDash
14. [x] CI running on AMD GPUs/CDash
15. [x] +single_memory experiment.py https://github.com/llnl/Kripke/pull/70
16. [x] +caliper experiment.py
17. [x] CPU-only performance analysis
18. [ ] CUDA performance analysis
19. [ ] ROCm performance analysis
20. [ ] Documentation (build, run instructions, performance demo)
21. [x] Add Umpire and force Chai to allocate before the main loop. Fixed in https://github.com/llnl/Kripke/pull/65
22. [x] Robert working on resolving race condition in LTimes https://github.com/llnl/Kripke/pull/71
23. [x] Problem size: 8x8 solve on each thread generates register pressure. How to pick the size based on memory?
24. [ ] Validation: Particle count, could have ground truth with some tolerance. An answer per problem size. Robert will check if all current platforms get the same answer and what the tolerance is.
25. [ ] single_memory variant in Chai: https://github.com/llnl/benchpark/pull/1373/changes#diff-bdf378bd4144585e9e2073fc11cdffadc308cf33fb38471b9e62b83009574191
26. [ ] Chai bug: bus error (Riyaz to give Allan a reproducer)
27. [ ] Verify C++ 20
28. [ ] Kripke implementation: Riyaz to show Allan
29. [ ] Thin chai allocation: no resource manager? Will not work on discrete memory GPU
30. [ ] Chai should not manage mesh data as it should always be on the GPU
31. [ ] Chai should only manage the MPI buffers
32. [ ] Chai always on, GPU aware MPI on or off