[mlir][Linalg] Avoid using `tensor.cast` by default while folding `fill` with `pad`.

[mravishankar]

Summary

This is unnecessary if the generated operation type already matches
the type of the replaced value. Also use OpFoldResult to reduce the
number of cases the casts are needed.

mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp

LogicalResult matchAndRewrite(tensor::PadOp padOp,
 
     ReifiedRankedShapedTypeDims reifiedShape;
     ReifyRankedShapedTypeOpInterface interface =
         cast<ReifyRankedShapedTypeOpInterface>(padOp.getOperation());
     if (failed(interface.reifyResultShapes(rewriter, reifiedShape)))
       return rewriter.notifyMatchFailure(
           padOp, "failed to reify tensor.pad op result shape");
 
-    auto oldResultType = padOp.getResultType();
-    SmallVector<int64_t, 4> staticShape(oldResultType.getRank(),
-                                        ShapedType::kDynamicSize);
+    SmallVector<OpFoldResult> newShape =
+        getAsOpFoldResult(reifiedShape.front());
     auto emptyTensor = rewriter.create<tensor::EmptyOp>(
-        padOp.getLoc(), staticShape, oldResultType.getElementType(),
-        reifiedShape.front());
-    auto newFillOp = rewriter.create<FillOp>(
-        fillOp.getLoc(), ValueRange{padValue}, ValueRange{emptyTensor});
-    rewriter.replaceOpWithNewOp<tensor::CastOp>(padOp, oldResultType,
-                                                newFillOp.result());
-
+        padOp.getLoc(), newShape, padOp.getResultType().getElementType());
+    Value replacement =
+        rewriter
+            .create<FillOp>(fillOp.getLoc(), ValueRange{padValue},
+                            ValueRange{emptyTensor})
+            .getResult(0);
+    if (replacement.getType() != padOp.getResultType()) {
+      replacement = rewriter.create<tensor::CastOp>(
+          fillOp.getLoc(), padOp.getResultType(), replacement);
+    }
+    rewriter.replaceOp(padOp, replacement);
     return success();
   }
 };
 
 /// Fold tensor.insert_slice(tensor.pad(<input>), linalg.fill) into
 /// tensor.insert_slice(<input>, linalg.fill) if the padding value and the
 /// filling value are the same.
 struct FoldInsertPadIntoFill : public OpRewritePattern<tensor::InsertSliceOp> {