[cuDNN] bump cuDNN-frontend submodule to 1.1.2 (#120761)
Hopefully resolves additional `CUDNN_STATUS_SUCCESS` failures that we have been seeing on H100 (though curiously not on upstream CI, perhaps due to the different hardware being tested)
Need to confirm the fix on our end before merging
CC @Skylion007
Pull Request resolved: https://github.com/pytorch/pytorch/pull/120761
Approved by: https://github.com/Skylion007, https://github.com/nWEIdia
