feat: add NumPy DeepPot

[njzjz]

While a DPModel cannot be directly trained, it can be converted from another model:

dp convert-backend frozen_model.pth frozen_model.dp
dp test -m frozen_model.dp -s ../data/

The energy result is consistent with TF and PT. Force and virial are NaN, as expected.

[codecov]

Codecov Report

Attention: Patch coverage is 69.01408% with 44 lines in your changes are missing coverage. Please review.

Project coverage is 75.39%. Comparing base (649fdca) to head (29737bc).

Files	Patch %	Lines
deepmd/dpmodel/infer/deep_eval.py	67.91%	43 Missing ⚠️
deepmd/dpmodel/utils/batch_size.py	83.33%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##            devel    #3332      +/-   ##
==========================================
- Coverage   75.41%   75.39%   -0.03%     
==========================================
  Files         406      408       +2     
  Lines       34106    34247     +141     
  Branches     1604     1604              
==========================================
+ Hits        25720    25819      +99     
- Misses       7525     7567      +42     
  Partials      861      861              

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