[mlir][Shape] Fix a crash when folding nary broadcast ops

[bkramer]

Summary

operands[2] can be nullptr here. I'm not able to build a lit test for
this because of the commutative reordering of operands. It's possible to
trigger this with a createOrFold<BroadcastOp> though.

[tpopp]

Whoops. That's definitely a better choice than what I did.

mlir/lib/Dialect/Shape/IR/Shape.cpp

 // BroadcastOp
 //===----------------------------------------------------------------------===//
 
 OpFoldResult BroadcastOp::fold(ArrayRef<Attribute> operands) {
   if (!operands[1])
     return nullptr;
 
   // TODO: Support folding with more than 2 input shapes
-  if (operands.size() > 2 && !operands[2].isa<StringAttr>())
+  if (shapes().size() > 2)
     return nullptr;
 
   auto rhsShape = llvm::to_vector<6>(
       operands[1].cast<DenseIntElementsAttr>().getValues<int64_t>());
   if (rhsShape.empty())
     return shapes()[0];
 
   if (!operands[0])