allow operator_range to take multiple ranges (#30124)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/30124
as title
Test Plan:
```
buck run mode/opt //caffe2/benchmarks/operator_benchmark:benchmark_all_other_test -- --tag_filter all --iterations 1 --device cuda --operator_range a,b-c
# ----------------------------------------
# PyTorch/Caffe2 Operator Micro-benchmarks
# ----------------------------------------
# Tag : all
# Benchmarking PyTorch: add
# Mode: Eager
# Name: add_M8_N32_K256_cuda
# Input: M: 8, N: 32, K: 256, device: cuda
Forward Execution Time (us) : 71.683
# Benchmarking PyTorch: batchnorm
# Mode: Eager
# Name: batchnorm_M1_N256_K3136_cuda
# Input: M: 1, N: 256, K: 3136, device: cuda
Forward Execution Time (us) : 118.840
# Benchmarking PyTorch: batchnorm
# Mode: Eager
# Name: batchnorm_M1_N8192_K1_cuda
# Input: M: 1, N: 8192, K: 1, device: cuda
Forward Execution Time (us) : 134.274
# Benchmarking PyTorch: cat
# Mode: Eager
# Name: cat_M128_N128_K1_dim1_cuda
# Input: M: 128, N: 128, K: 1, dim: 1, device: cuda
Forward Execution Time (us) : 109.172
...
Reviewed By: hl475
Differential Revision: D18605640
fbshipit-source-id: 4ae9b91a50c4cdf1b161b6c5c58f365ba514050c
