Serialization and deserialization of kernel

[ikkoham]

Required prerequisites

• Search the issue tracker to check if your feature has already been mentioned or rejected in other issues.

Describe the feature

The ability to save and load the kernel is needed.

The use case is, for example, to save kernels, used as a cache, or broadcast with MPI.
There are many ways to implement this: QASM3, some MLIR, or even the original format.

Requirements

• not only Python but also C++
• Backward compatibility, i.e., you can load a saved kernel from a future version

Reference: CUDA has JIT feature and it can be stored.

• https://docs.nvidia.com/cuda/nvrtc/index.html
• https://github.com/NVIDIA/jitify

Deserialization is related to #755