Simplify logic for typing ASSIGN nodes in TypeCheck

There's some existing logic that will put the rhs type on an ASSIGN node iff assign is valid, or the unknown type otherwise.

This is pointless now because we already put the rhs type on an ASSIGN unconditionally in TypeInference. This case only triggers for code that (for whatever reason) doesn't get typed during TypeInference.

-------------
Created by MOE: https://github.com/google/moe
MOE_MIGRATED_REVID=207810074

src/com/google/javascript/jscomp/TypeCheck.java

@@ -1074,7 +1074,9 @@ private void visitAssign(NodeTraversal t, Node assign) {
       }
     }
 
-    checkCanAssignToWithScope(t, assign, lvalue, getJSType(rvalue), "assignment");
+    JSType rightType = getJSType(rvalue);
+    checkCanAssignToWithScope(t, assign, lvalue, rightType, "assignment");
+    ensureTyped(assign, rightType);
   }
 
   /**
@@ -1115,14 +1117,9 @@ private void checkCanAssignToWithScope(
           leftType = var.getType();
         }
       }
-    }
+    } // Fall through case for arbitrary LHS and arbitrary RHS.
 
-    // Fall through case for arbitrary LHS and arbitrary RHS.
-    if (validator.expectCanAssignTo(t, assign, rightType, leftType, msg)) {
-      ensureTyped(assign, rightType);
-    } else {
-      ensureTyped(assign);
-    }
+    validator.expectCanAssignTo(t, assign, rightType, leftType, msg);
   }
 
   private void checkPropCreation(NodeTraversal t, Node lvalue) {