Add dpgui entry point and dpgen gui CLI

[njzjz]

See deepmodeling/dpgui#270 for details.

[codecov]

Codecov Report

Attention: 12 lines in your changes are missing coverage. Please review.

Comparison is base (8c47020) 48.64% compared to head (d6ee1e0) 48.62%.

❗ Current head d6ee1e0 differs from pull request most recent head 0851f48. Consider uploading reports for the commit 0851f48 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##            devel    #1349      +/-   ##
==========================================
- Coverage   48.64%   48.62%   -0.02%     
==========================================
  Files          82       83       +1     
  Lines       14681    14693      +12     
==========================================
+ Hits         7141     7144       +3     
- Misses       7540     7549       +9     

Files	Coverage Δ
dpgen/main.py	0.00% <0.00%> (ø)
dpgen/gui.py	0.00% <0.00%> (ø)

... and 1 file with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[njzjz]

There seems to be an issue here. Python will dump NaN as {"type": NaN}. However, JavaScript doesn't think it a valid JSON string.

Python:

>>> import json
>>> json.dumps({"type":float("nan")})
'{"type": NaN}'

JavaScript:

> JSON.parse('{"type": NaN}')
VM1054:1 Uncaught SyntaxError: Unexpected token 'N', "{"type": NaN}" is not valid JSON
    at JSON.parse (<anonymous>)
    at <anonymous>:1:6

json.dumps(https://docs.python.org/3/library/json.html#json.dumps) has a allow_nan argument, but when it is False, an error is thrown.

[njzjz]

Python will dump NaN as {"type": NaN}.

Finally, I changed NaN to 1e10 to avoid this issue.