[Fix] Fix onnx2ncnn.cpp bugs.

[hanrui1sensetime]

Thanks for your contribution and we appreciate it a lot. The following instructions would make your pull request more healthy and more easily receiving feedbacks. If you do not understand some items, don't worry, just make the pull request and seek help from maintainers.

Motivation

Fix some of onnx2ncnn.cpp bugs. Especially for RTMPose model deployment later.

Modification

-[x] Fix Gemm for opset > 9, which is optional for the third input, while original onnx2ncnn needs the third input strictly.
-[x] Sync the Reduction op from the newest onnx2ncnn in ncnn

Other ops sync from the newest onnx2ncnn in ncnn
will be fixed later in another PR. This PR is mainly for RTMPose model.

BC-breaking (Optional)

None.

Use cases (Optional)

None.

Checklist

1. Pre-commit or other linting tools are used to fix the potential lint issues.
2. The modification is covered by complete unit tests. If not, please add more unit tests to ensure the correctness.
3. If the modification has a dependency on downstream projects of a newer version, this PR should be tested with all supported versions of downstream projects.
4. The documentation has been modified accordingly, like docstring or example tutorials.

[CLAassistant]

All committers have signed the CLA.

[RunningLeon]

could add ut for this case

[lvhan028]

could add ut for this case

It's onnx2ncnn.cpp, coming from ncnn repo. I think we don't need to add unit tests for it.

[RunningLeon]

LGTM