Skip failing test_torchinductor_codegen_dynamic_shapes tests on CPU (#98621)
This test starts to fail in trunk after https://github.com/pytorch/pytorch/pull/97230. The original PR missed this because these test are marked as slow and is only run periodically. Is this ok to skip them like `test_upsample_cat_conv_dynamic_shapes`?
Here is an example failure https://github.com/pytorch/pytorch/actions/runs/4638277468/jobs/8208270657. The following tests are all failing with `Failed to find dynamic for loop variable` error like others in the list. They are:
* `test_conv2d_binary_dynamic_shapes`. Fixes https://github.com/pytorch/pytorch/issues/98679
* `test_conv2d_unary_dynamic_shapes`. Fixes https://github.com/pytorch/pytorch/issues/98680
* `test_conv_bn_fuse_dynamic_shapes`. Fixes https://github.com/pytorch/pytorch/issues/98681
* `test_conv_transpose2d_unary_dynamic_shapes`. Fixes https://github.com/pytorch/pytorch/issues/98682
Pull Request resolved: https://github.com/pytorch/pytorch/pull/98621
Approved by: https://github.com/malfet
