[clang] [mlir] [CIR] Implement 'allocsize' function/call attribute lowering (PR #179342)

[Tobias Gysi via cfe-commits]

================
@@ -2725,6 +2726,23 @@ static void convertNoBuiltinAttrs(MLIRContext *ctx,
     target.setNobuiltinsAttr(ArrayAttr::get(ctx, nbAttrs.getArrayRef()));
 }
 
+template <typename OpTy>
+static void convertAllocsizeAttr(MLIRContext *ctx,
+                                 const llvm::AttributeSet &attrs, OpTy target) {
+  llvm::Attribute attr = attrs.getAttribute(llvm::Attribute::AllocSize);
+  if (!attr.isValid())
+    return;
+
+  auto [elemSize, numElems] = attr.getAllocSizeArgs();
+  if (numElems)
+    target.setAllocsizeAttr(
+        DenseI32ArrayAttr::get(ctx, {static_cast<int32_t>(elemSize),
+                                     static_cast<int32_t>(*numElems)}));
+  else
+    target.setAllocsizeAttr(
+        DenseI32ArrayAttr::get(ctx, {static_cast<int32_t>(elemSize)}));
----------------
gysit wrote:

```suggestion
  if (numElems) {
    target.setAllocsizeAttr(
        DenseI32ArrayAttr::get(ctx, {static_cast<int32_t>(elemSize),
                                     static_cast<int32_t>(*numElems)}));
  } else {
    target.setAllocsizeAttr(
        DenseI32ArrayAttr::get(ctx, {static_cast<int32_t>(elemSize)}));
  }
```
ultra nit: I would go with braces here.

https://github.com/llvm/llvm-project/pull/179342