This PR adds two utility functions to allow us to profile a function
with nsys/ncu in the code.

Usage:

```python
import os
import pytest
import torch
from hidet.utils.nsys_utils import nsys_run, nsys_get_path

def func(b):
    a = torch.randn(1000, device='cuda')
    a + b


def main():
    # profile with nsight system, store the result to 'nsys-reports/...'
    report = nsys_run(func, b=1)

    # we can visualize the profiling result by calling the `visualize` method.
    report.visualize()

if __name__ == '__main__':
    # we need to wrap the call to `main` inside '__main__' so that this script can be "imported" without 
    # side effect, which is one requirement for the utility
    main()
```

This PR adds two utility functions to allow us to profile a function
with nsys/ncu in the code.

Usage:

```python
import os
import pytest
import torch
from hidet.utils.nsys_utils import nsys_run, nsys_get_path

def func(b):
    a = torch.randn(1000, device='cuda')
    a + b


def main():
    # profile with nsight system, store the result to 'nsys-reports/...'
    report = nsys_run(func, b=1)

    # we can visualize the profiling result by calling the `visualize` method.
    report.visualize()

if __name__ == '__main__':
    # we need to wrap the call to `main` inside '__main__' so that this script can be "imported" without 
    # side effect, which is one requirement for the utility
    main()
```

This PR adds two utility functions to allow us to profile a function
with nsys/ncu in the code.

Usage:

```python
import os
import pytest
import torch
from hidet.utils.nsys_utils import nsys_run, nsys_get_path

def func(b):
    a = torch.randn(1000, device='cuda')
    a + b


def main():
    # profile with nsight system, store the result to 'nsys-reports/...'
    report = nsys_run(func, b=1)

    # we can visualize the profiling result by calling the `visualize` method.
    report.visualize()

if __name__ == '__main__':
    # we need to wrap the call to `main` inside '__main__' so that this script can be "imported" without 
    # side effect, which is one requirement for the utility
    main()
```