[rocm6.1] Update Triton dependency logic

[jithunnair-amd]

• Reinstate PyTorch dependency on hashed version of triton (essentially reverts e725944): Tested successfully via http://rocm-ci.amd.com/job/rocm-pytorch-manylinux-wheel-builder/154
• Do not add triton dependency for PyTorch versions older than 1.13: Tested successfully via http://rocm-ci.amd.com/job/rocm-pytorch-manylinux-wheel-builder/155

[jithunnair-amd]

Test results of PyTorch2.0 and triton wheels generated from http://rocm-ci.amd.com/job/rocm-pytorch-manylinux-wheel-builder/154, installed in rocm/pytorch-private:manylinux-hipclang-13484:

[root@ed0329edef16 apex]# pip3 install -f https://compute-artifactory.amd.com/artifactory/compute-pytorch-rocm/compute-rocm-dkms-no-npi-hipclang/13484/ torch==2.0.1 torchvision==0.15.2
Looking in links: https://compute-artifactory.amd.com/artifactory/compute-pytorch-rocm/compute-rocm-dkms-no-npi-hipclang/13484/
Collecting torch==2.0.1
  Downloading https://compute-artifactory.amd.com/artifactory/compute-pytorch-rocm/compute-rocm-dkms-no-npi-hipclang/13484/torch-2.0.1%2Brocm6.1-cp39-cp39-linux_x86_64.whl (2095.9 MB)
...
Collecting pytorch-triton-rocm==2.0.2+f84c1f1e62 (from torch==2.0.1)
  Downloading https://compute-artifactory.amd.com/artifactory/compute-pytorch-rocm/compute-rocm-dkms-no-npi-hipclang/13484/pytorch_triton_rocm-2.0.2%2Bf84c1f1e62-cp39-cp39-linux_x86_64.whl (206.3 MB)
...

[root@ed0329edef16 pytorch-micro-benchmarking]# python3 micro_benchmarking_pytorch.py --network resnet50 --compile
INFO: running forward and backward for warmup.
/opt/python/cp39-cp39/lib/python3.9/site-packages/torch/_inductor/compile_fx.py:90: UserWarning: TensorFloat32 tensor cores for float32 matrix multiplication available but not enabled. Consider setting `torch.set_float32_matmul_precision('high')` for better performance.
  warnings.warn(
INFO: running the benchmark..
OK: finished running benchmark..
--------------------SUMMARY--------------------------
Microbenchmark for network : resnet50
Num devices: 1
Dtype: FP32
Mini batch size [img] : 64
Time per mini-batch : 0.11212306022644043
Throughput [img/sec] : 570.8014022338267