Support deterministic algorithm configuration in PyTorch backend

[yhna940]

Is your feature request related to a problem? Please describe.
Some users require deterministic outputs for reproducibility, but the PyTorch backend does not currently expose a way to enable deterministic algorithms.

Describe the solution you'd like
Add a model configuration parameter to enable at::Context::setDeterministicAlgorithms(...) during model execution.

Describe alternatives you've considered
N/A

Additional context
I’ve opened a small PR to add this feature. Would appreciate a review: triton-inference-server/pytorch_backend#150