Fix inductor CI (by updating graph break count) (pytorch#110160)

There was a vision hash update which led to fewer graph breaks. This
seems expected to me (because the hash update included
pytorch/vision#7944 and nms is used in maskrcnn).

Test Plan:
- wait for ci

Pull Request resolved: pytorch#110160
Approved by: https://github.com/ezyang, https://github.com/Chillee

benchmarks/dynamo/ci_expected_accuracy/inductor_torchbench_training.csv

@@ -50,5 +50,5 @@ timm_vision_transformer_large,pass_due_to_skip,0
 timm_vovnet,pass,8
 tts_angular,pass,10
 vgg16,pass,8
-vision_maskrcnn,fail_accuracy,42
+vision_maskrcnn,fail_accuracy,39
 yolov3,pass,10