Skip testing triangular_solve_batched on non-default CUDA stream (#26115)
Summary:
This is for testing purposes.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/26115
Differential Revision: D17433122
Pulled By: zou3519
fbshipit-source-id: bf41327e6141e9ae589fcf18254c2a8cdd868dd7
