[Feature] Support PAConv operation

[Wuziyi616]

Implement PAConv layer. PAConv is similar to Conv2d or MLP, serving as basic building block of a model. The basic idea of PAConv is simple. PAConv dynamically generates convolutional kernels by assembling some weight kernels in its weight bank. Every PAConv has a weight bank which consists of several (typically 16) trainable weight kernels, and a ScoreNet which is implemented by MLPs. Given a point pair p_c is a center and p_1 is its neighbor (queried by KNN), it uses ScoreNet to predict scores for this point pair, and uses the score to assemble weight bank to form one kernel. Then, it runs convolution using the assembled kernel on the point pair and get output features.

PAConv has two versions of implementation, one is called PAConv and the other one is called PAConvCUDA. The differences are:

• PAConv is easy to understand, and can run on both CPU as well as GPU. Its inputs are features already grouped by KNN, so it only needs to compute scores and assemble weights and get the output.
• PAConvCUDA is tricky, and can only run on GPU because it utilizes a custom cuda op assign_score_withk. Its inputs are features NOT grouped by KNN, and queries KNN on the fly (hard to describe). This saves memory because we don't need to pre-compute KNN features which have very large tensor size (typically (B, C, npoint, K)).

The operation is somewhat complex and I am not sure if I should write its detailed process in the comment. Also, the original code is dirty, although I tried my best to add comments and made it clear, it may still not meet our standard. Please kindly help me improve it (e.g. some function's name).

[codecov]

Codecov Report

Merging #598 (cd8730e) into master (20c0fbd) will increase coverage by 0.07%.
The diff coverage is 57.22%.

@@            Coverage Diff             @@
##           master     #598      +/-   ##
==========================================
+ Coverage   50.92%   51.00%   +0.07%     
==========================================
  Files         197      201       +4     
  Lines       15056    15220     +164     
  Branches     2444     2467      +23     
==========================================
+ Hits         7668     7763      +95     
- Misses       6884     6939      +55     
- Partials      504      518      +14     

Flag	Coverage Δ
unittests	51.00% <57.22%> (+0.07%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
mmdet3d/ops/group_points/group_points.py	31.11% <12.50%> (-1.04%)	⬇️
mmdet3d/ops/paconv/assign_score.py	29.16% <29.16%> (ø)
mmdet3d/ops/paconv/utils.py	52.38% <52.38%> (ø)
mmdet3d/ops/paconv/paconv.py	66.05% <66.05%> (ø)
mmdet3d/ops/__init__.py	100.00% <100.00%> (ø)
mmdet3d/ops/paconv/__init__.py	100.00% <100.00%> (ø)

---

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 20c0fbd...cd8730e. Read the comment docs.

[Wuziyi616]

Would be better if you read the paper before reviewing my code.

[Wuziyi616]

I think this PR is ready for review @Tai-Wang @ZwwWayne