[distributed] Configure collective timeouts through ProcessGroup (#193238)

Distributed tests configured timeouts through a private backend method, which required nccl2 to expose an unnecessary compatibility alias. Use the public ProcessGroup timeout API instead so timeout configuration follows the common c10d path.

Test Plan:

```bash
CUDA_VISIBLE_DEVICES=0,1 VIRTUAL_ENV=/home/tristanr/pytorch/.venv PATH=/home/tristanr/pytorch/.venv/bin:$PATH /home/tristanr/.local/bin/uv run --no-project test/distributed/test_c10d_nccl.py ProcessGroupNCCLGroupTest -k test_set_nccl_pg_timeout
CUDA_VISIBLE_DEVICES=0,1 VIRTUAL_ENV=/home/tristanr/pytorch/.venv PATH=/home/tristanr/pytorch/.venv/bin:$PATH /home/tristanr/.local/bin/uv run --no-project test/distributed/test_c10d_nccl.py ProcessGroupNCCLGroupTest -k test_extend_nccl_pg_timeout
```

Authored with an AI assistant.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/193238
Approved by: https://github.com/kapilsh
ghstack dependencies: #193237