add support for hdf5

[njzjz]

dpdata format will be provided later.

The advantage of hdf5 is that it only has one file for multi-systems containing many systems, so it will be easier to transfer between servers or share online.

File size

About the file size, take an example of methane_v1:

• Original file: 320 MB
• Compressed file (.tar.bz2): 182 MB
• HDF5 file: 258 MB
• Compressed HDF5 file (.tar.bz2): 183 MB

We can see HDF5 file will have no advantage if fully compressed, but it is smaller if not compressed. Also, compressing and decompressing a single file should be faster than compressing a lot of small files.

Usage

One can add a HDF5 file in the systems list in input JSON file.

[codecov-commenter]

Codecov Report

Merging #1163 (000c768) into devel (077df3d) will decrease coverage by 11.38%.
The diff coverage is n/a.

❗ Current head 000c768 differs from pull request most recent head dc44906. Consider uploading reports for the commit dc44906 to get more accurate results

@@             Coverage Diff             @@
##            devel    #1163       +/-   ##
===========================================
- Coverage   75.66%   64.28%   -11.39%     
===========================================
  Files          89        5       -84     
  Lines        7048       14     -7034     
===========================================
- Hits         5333        9     -5324     
+ Misses       1715        5     -1710     

Impacted Files	Coverage Δ
deepmd/entrypoints/config.py
deepmd/descriptor/__init__.py
deepmd/infer/deep_polar.py
deepmd/entrypoints/__init__.py
deepmd/descriptor/se_a.py
deepmd/infer/deep_eval.py
deepmd/entrypoints/doc.py
deepmd/utils/learning_rate.py
deepmd/descriptor/se.py
deepmd/__main__.py
... and 71 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 077df3d...dc44906. Read the comment docs.