Update on "[lang] Record the element_type of the AnyArray"

<!--
copilot:all
-->
### <samp>🤖 Generated by Copilot at 0176872</samp>

### Summary
📝🔬🆕

<!--
1.  📝 - This emoji can represent the change of modifying the `decl_ndarray_arg` function to pass the `element_type` to the `AnyArray` constructor, since this is a code change that involves writing or editing some code.
2.  🔬 - This emoji can represent the change of adding `element_type` argument to `AnyArray` constructor and using it in `get_type` and `grad` methods, since this is a code change that involves improving the type handling and gradient computation of arbitrary arrays, which are related to scientific or mathematical operations.
3.  🆕 - This emoji can represent the change of introducing a new argument to the `AnyArray` constructor, since this is a code change that involves adding a new feature or functionality to the existing code.
-->
Improved type handling and gradient computation of `AnyArray` arguments in Taichi kernels and functions. Added `element_type` parameter to `AnyArray` constructor and `decl_ndarray_arg` function.

> _To declare an array argument_
> _We pass the element type along_
> _This helps `AnyArray`_
> _To know what to say_
> _When it calls `get_type` or `grad`_

### Walkthrough
*  Add `element_type` argument to `AnyArray` constructor and store it as an attribute ([link](https://github.com/taichi-dev/taichi/pull/8192/files?diff=unified&w=0#diff-2e623ee0b0eec1b200fead36c0627a3c54738f6d83d79757398dc67decc01da8L18-R20), [link](https://github.com/taichi-dev/taichi/pull/8192/files?diff=unified&w=0#diff-575efc738df7b1202370c2531ec82232dc7f287b2bec4999af03ef40da4f5deeL111-R111))
*  Return stored `element_type` in `AnyArray.get_type` method instead of inferring from pointer ([link](https://github.com/taichi-dev/taichi/pull/8192/files?diff=unified&w=0#diff-2e623ee0b0eec1b200fead36c0627a3c54738f6d83d79757398dc67decc01da8L36-R37))
*  Pass stored `element_type` to `AnyArray` constructor in `AnyArray.grad` method to preserve type information in gradient array ([link](https://github.com/taichi-dev/taichi/pull/8192/files?diff=unified&w=0#diff-2e623ee0b0eec1b200fead36c0627a3c54738f6d83d79757398dc67decc01da8L43-R44))






[ghstack-poisoned]

taichi/ir/frontend_ir.cpp

@@ -1677,20 +1677,22 @@ std::vector<Expr> ASTBuilder::expand_exprs(const std::vector<Expr> &exprs) {
         expanded_exprs.push_back(expr);
       } else {
         // Expand TensorType expr
+        // clang-format off
         /*
           Before:
             TensorType<4 x i32> index = Expr;
 
-      After:
-        TensorType<4 x i32>* id_expr = FrontendAllocaStmt(TensorType<4 x i32>)
+          After:
+            TensorType<4 x i32>* id_expr = FrontendAllocaStmt(TensorType<4 x i32>)
             i32 ind0 = IndexExpression(id_expr, 0)
-                i32 ind1 = IndexExpression(id_expr, 1)
-                    i32 ind2 = IndexExpression(id_expr, 2)
-                        i32 ind3 = IndexExpression(id_expr, 3)
+            i32 ind1 = IndexExpression(id_expr, 1)
+            i32 ind2 = IndexExpression(id_expr, 2)
+            i32 ind3 = IndexExpression(id_expr, 3)
 
-                                       return {ind0, ind1, ind2, ind3}
+            return {ind0, ind1, ind2, ind3}
 
-                                           */
+         */
+        // clang-format on
         auto tensor_type = expr->ret_type.ptr_removed()->cast<TensorType>();
 
         Expr id_expr;