[type] Add BitStructStoreStmt and corresponding passes

[yuanming-hu]

Related issue = #1905

benchmark_bit_struct_stores.py:

• CPU: 805ms -> 271ms
• CUDA: 13.6ms -> 6.73 ms

[Click here for the format server]

---

[xumingkuan]

I'm not sure how will BitStructStoreStmt affect data_sourse_analysis.cpp and other potential IR optimizations assuming GlobalStoreStmt and AtomicOpStmt are the only statements modifying global fields. However, we are good now as long as this pass is after all other optimizations.

[yuanming-hu]

I'm not sure how will BitStructStoreStmt affect data_sourse_analysis.cpp and other potential IR optimizations assuming GlobalStoreStmt and AtomicOpStmt are the only statements modifying global fields. However, we are good now as long as this pass is after all other optimizations.

Exactly. It seems that we need clear separations between the optimization passes in the future...