fix print tensor in cpp for privateuse1 (#100797)
Fixes #ISSUE_NUMBER
1、fix lintrunnr in `test/inductor/test_cuda_repro.py`
2、In Libtorch, if we rename the `privateuseone` backend to `foo`, and when we print tensor with `std::cout << tensor`, we will get the output like this,
```
1.0, 2.0 ...
[PrivateUse1FloatType{2,3}]
```
and it should be like this
```
1.0, 2.0 ...
[fooFloatType{2,3}]
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/100797
Approved by: https://github.com/ezyang
