make the pairwise DPRc model 2x faster

[njzjz]

This PR does a trick to speed up the pairwise DPRc model. Considering #2618 is not ready and is quite difficult to implement, in this PR, multiple frames are merged into one frame before feed to prod_env_mat OP, and the mesh is faked to make it perform the same behavior as the multiple frames.
A new mesh shape is proposed. The first element stores nloc, and the following 15 elements store nothing to distinguish it from other mesh. The (16 : 16 + nloc) elements store ilist, (16 + nloc : 16 + nloc * 2) store numneigh, and the rest elements (in the shape of sum(numneigh)) store neighbors. The nei_mode is 4 for this situation.

prod_env_mat OP is not a bottleneck anymore, as shown below.

[codecov]

Codecov Report

Patch coverage: 82.63% and project coverage change: +0.06% 🎉

Comparison is base (0d5737f) 75.45% compared to head (f25cbb6) 75.52%.
Report is 6 commits behind head on devel.

Additional details and impacted files

@@            Coverage Diff             @@
##            devel    #2833      +/-   ##
==========================================
+ Coverage   75.45%   75.52%   +0.06%     
==========================================
  Files         242      242              
  Lines       24205    24370     +165     
  Branches     1537     1571      +34     
==========================================
+ Hits        18265    18405     +140     
- Misses       5023     5037      +14     
- Partials      917      928      +11     

Files Changed	Coverage Δ
source/op/prod_env_mat_multi_device.cc	70.96% <57.89%> (-0.40%)	⬇️
deepmd/model/pairwise_dprc.py	93.49% <80.00%> (-3.00%)	⬇️
source/op/pairwise.cc	85.95% <87.61%> (+1.34%)	⬆️

... and 3 files with indirect coverage changes

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