[Sparse] Cannot find mkl libraries when linking sparse library against conda installed Pytorch.

## 🐛 Bug

Current DGL sparse library will link against libtorch. It works well when we use pip to install pytorch. However, if we install pytorch with conda, it has mkl dependency by default, leading to compilation error if mkl is not installed locally. A relevant issue in Pytorch is https://github.com/pytorch/pytorch/issues/73008.

## To Reproduce

Steps to reproduce the behavior:

1.
1.
1.

## Expected behavior

## Environment

- DGL Version (e.g., 1.0):
- Backend Library & Version (e.g., PyTorch 0.4.1, MXNet/Gluon 1.3):
- OS (e.g., Linux):
- How you installed DGL (`conda`, `pip`, source):
- Build command you used (if compiling from source):
- Python version:
- CUDA/cuDNN version (if applicable):
- GPU models and configuration (e.g. V100):
- Any other relevant information:

## Additional context