[Feature] NMS update

[SemyonBevzuk]

Added support for "score_threshold" and "max_num" for NMS operation when exporting to ONNX.
These parameters are supported in ONNX: link.

These changes significantly speed up the export time:

Model name	Export time before (s)	Export time now (s)
ssd_300	1201.27	6.85
yolo_v3	88.89	73.11
fsaf	19.89	12.62
retinanet	59.33	13.30
faster_rcnn	21.24	18.02

In addition, performance is improved in ONNX Runtime.
This can be seen in the example of the RetinaNet model.
Performance measured using standard ONNX Runtime tools on 10 random inputs.
The table shows the average execution time of the NMS operation and the total inference time of the model:

RetinaNet (ONNX Runtime)	Befor (s)	Now (s)
NMS time	15.99	0.0002574
Total time	16.78	0.76766

[CLAassistant]

All committers have signed the CLA.

[codecov]

Codecov Report

Merging #957 (85576ea) into master (4d42365) will increase coverage by 0.49%.
The diff coverage is 79.64%.

@@            Coverage Diff             @@
##           master     #957      +/-   ##
==========================================
+ Coverage   65.20%   65.70%   +0.49%     
==========================================
  Files         156      157       +1     
  Lines       10034    10086      +52     
  Branches     1816     1829      +13     
==========================================
+ Hits         6543     6627      +84     
+ Misses       3154     3113      -41     
- Partials      337      346       +9     

Flag	Coverage Δ
unittests	65.70% <79.64%> (+0.49%)	⬆️

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

Impacted Files	Coverage Δ
mmcv/tensorrt/tensorrt_utils.py	0.00% <0.00%> (-2.39%)	⬇️
mmcv/tensorrt/__init__.py	60.00% <57.14%> (+10.00%)	⬆️
mmcv/tensorrt/preprocess.py	78.66% <78.66%> (ø)
mmcv/ops/nms.py	61.63% <100.00%> (+5.20%)	⬆️
mmcv/onnx/info.py	68.75% <0.00%> (-6.25%)	⬇️
mmcv/onnx/onnx_utils/symbolic_helper.py	35.46% <0.00%> (+1.74%)	⬆️
mmcv/tensorrt/init_plugins.py	50.00% <0.00%> (+6.25%)	⬆️

---

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 4d42365...85576ea. Read the comment docs.

[RunningLeon]

@SemyonBevzuk , Could you fix CI error?

[RunningLeon]

@ZwwWayne See if you have any comments on this PR.

[RunningLeon]

@SemyonBevzuk hi，could you kindly fix the CI error?

[RunningLeon]

LGTM

[RunningLeon]

Kindly ping @ZwwWayne