[perf] Fix Taichi CPU backend compile parameter to pair performance w…

…ith Numba. (#7731)

Issue: #7442

### Brief Summary

In this issue, Numba is a magnitude faster than Taichi due to the
absence of automatic vectorization.
The root cause is the incorrect passage of the `fast_flag`.

To solve this problem, `fast_flag` is now added to the initialization of
cpu codegen. Numba and Taichi now reveal comparable performance.
Here's perf comparison:
numba:            13052.542478MFlops
taichi(master): 6544.274409MFlops
taichi(this pr):  12778.240179MFlops

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

taichi/codegen/llvm/codegen_llvm.cpp

@@ -2542,6 +2542,13 @@ void TaskCodeGenLLVM::initialize_context() {
   TI_ASSERT(tlctx != nullptr);
   llvm_context = tlctx->get_this_thread_context();
   builder = std::make_unique<llvm::IRBuilder<>>(*llvm_context);
+  if (compile_config.fast_math) {
+    llvm::FastMathFlags fast_flags;
+    fast_flags.setNoInfs();
+    fast_flags.setNoSignedZeros();
+    fast_flags.setAllowReassoc();
+    builder->setFastMathFlags(fast_flags);
+  }
 }
 
 llvm::Value *TaskCodeGenLLVM::get_arg(int i) {