[aot] Support push_arg for Kernel in C++ wrapper (taichi-dev#6419)

fixes taichi-dev#6413 

### Brief Summary
Note we also changed the way to set vec/matrix args the previous `set`
didn't save users from counting index manually. cc: @YuCrazing

c_api/include/taichi/cpp/taichi.hpp

@@ -507,17 +507,27 @@ class Kernel {
     return at(i);
   }
 
-  // Temporary workaround for setting vec/matrix arguments in a flattened way.
   template <typename T>
-  void set(uint32_t i, const std::vector<T> &v) {
-    if (i + v.size() >= args_.size()) {
-      args_.resize(i + v.size());
-    }
+  void push_arg(const std::vector<T> &v) {
+    int idx = args_.size();
+    // Temporary workaround for setting vec/matrix arguments in a flattened way.
+    args_.resize(args_.size() + v.size());
     for (int j = 0; j < v.size(); ++j) {
-      at(i + j) = v[j];
+      at(idx + j) = v[j];
     }
   }
 
+  template <typename T>
+  void push_arg(const T &arg) {
+    int idx = args_.size();
+    args_.resize(idx + 1);
+    at(idx) = arg;
+  }
+
+  void clear_args() {
+    args_.clear();
+  }
+
   void launch(uint32_t argument_count, const TiArgument *arguments) {
     ti_launch_kernel(runtime_, kernel_, argument_count, arguments);
   }

c_api/tests/c_api_aot_test.cpp

@@ -21,9 +21,13 @@ static void kernel_aot_test(TiArch arch) {
 
   std::vector<int> arg2_v = {1, 2, 3};
 
-  k_run[0] = arg0_val;
-  k_run[1] = arg1_array;
-  k_run.set(2, arg2_v);
+  // This is just to make sure clear_args() does its work.
+  k_run.push_arg(arg0_val);
+  k_run.clear_args();
+
+  k_run.push_arg(arg0_val);
+  k_run.push_arg(arg1_array);
+  k_run.push_arg(arg2_v);
   k_run.launch();
   runtime.wait();