[Feature] Support CIHP dataset

[linfangjian01]

Thanks for your contribution and we appreciate it a lot. The following instructions would make your pull request more healthy and more easily get feedback. If you do not understand some items, don't worry, just make the pull request and seek help from maintainers.

Motivation

CIHP is tasked for multiple human parsing. There are 19 semantic classes and 1 background class. The dataset is divided into 28K/5K/5K images for training, validation and testing.

Modification

1. Add mmseg/datasets/cihp.py to get the images and labels.
2. Add CIHPDataset in mmseg/datasets/init.py.
3. Add configs/base/datasets/cihp.py for CIHP dataset.
4. Add deeplabv3plus and pspnet config files of CIHP dataset.