VM: Fix a couple of issues in the AOT optimizer, add fast path smi multiply.

runtime/vm/intermediate_language_arm64.cc

Index: runtime/vm/intermediate_language_arm64.cc
diff --git a/runtime/vm/intermediate_language_arm64.cc b/runtime/vm/intermediate_language_arm64.cc
index cb225a2ac6b79183f2aa8962af8c23bf21bc4ba3..5bd6acbcc9550790b590c74cc1ac832023ab6e05 100644
--- a/runtime/vm/intermediate_language_arm64.cc
+++ b/runtime/vm/intermediate_language_arm64.cc
@@ -2821,6 +2821,14 @@ void CheckedSmiOpInstr::EmitNativeCode(FlowGraphCompiler* compiler) {
       __ subs(result, left, Operand(right));
       __ b(slow_path->entry_label(), VS);
       break;
+    case Token::kMUL:
+      __ SmiUntag(TMP, left);
+      __ mul(result, TMP, right);
+      __ smulh(TMP, TMP, right);
+      // TMP: result bits 64..127.
+      __ cmp(TMP, Operand(result, ASR, 63));
+      __ b(slow_path->entry_label(), NE);
+      break;
     case Token::kBIT_OR:
       // Operation part of combined smi check.
       break;