[Simulator][Dataflow] Support parallel simulation for kernel with independent PEs

[Oh-Afterglow]

Description

This PR enables parallel simulation for kernels with independent PEs, i.e. no FIFO communications between PEs. It's provided as a backend LLVMOMPModule including 2 parts:

• A transformation pass that injects operations of the OpenMP dialect (omp) into kernel's MLIR code, which separates PE function calls into different omp.sections and enables parallelism among different PEs.
• A lowering pipeline that lowers the transformed MLIR code into the simulator executable for CPU (a mixture of llvm and omp dialect operations able to run on the MLIR execution engine).

Examples

The test case of tiled GEMM is provided.

Checklist

• PR's title starts with a category (e.g. [Bugfix], [IR], [Builder], etc)
• Changes are complete (i.e. I finished coding on this PR)
• All changes have test coverage (It would be better to provide ~2 different test cases to test the robustness of your code)
• Code is well-documented

[chhzh123]

Thanks for submitting a PR. Since we are building a simulator which is supposed to be different from the original LLVM backend, it'd be better to separate the implementation from the original one. For example, you can create a new file called simulator.py under allo/backend and specify the target as "simulator" instead of overwriting "llvm".

[Oh-Afterglow]

I've restructured the files as requested.