AOT for the LLVM backend

[k-ye]

• Make CPU backend work
  • Support arg-less kernels
  • Support scalar + Ndarray args
  • Split LlvmProgramImpl into compile-time & runtime parts to form Runtime API
• Make CUDA backend work
  • Support arg-less kernels
  • Support scalar + Ndarray args
• Provide public interfaces
  • Interface with Compute Graph
  • Interface with C-API
• Improvements & Demo
  • Dense/Pointer SNodes Demo
  • Support more flexible snode_id, snode_tree_id (determined at runtime)
  • Implement SNodeTreeType: https://github.com/taichi-dev/taichi/blob/master/docs/rfcs/20220413-aot-for-all-snode.md
  • Support memory deallocation for CPU backend. Merge Device::allocate_memory_runtime() with Device::allocate_memory()
  • Verify SNode GC works for Field AOT.
  • Serialize/deserialize Ndarray/SNodeTree buffers: Serialize and deserialize sparse data structrure from/to buffer and file #5428