Support Parallel HPO strategy for NNI-Tuun

**Describe the new feature and the current behavior/state**
Currently, Tuun does not support parallel parameter optimization under the NNI setting. It would be helpful to integrate this feature.

**Will this change the current API? How?**
N/A

**Describe alternatives you've considered**
N/A

**Additional context**
N/A

