[NOT Merge] add new model example pr

[huangjun12]

This pr just shows how to add new model into paddlevideo.
more details can be refer to add_new_algorithm doc

Try to run, just

git clone  https://github.com/huangjun12/PaddleVideo.git
git checkout contribute_pr
cd PaddleVideo
bash run.sh

log:

[11/26 12:27:46] epoch:[  1/5  ] train step:0    loss: 2.29299 lr: 0.001000 top1: 0.20000 top5: 0.60000 batch_cost: 0.90697 sec, reader_cost: 0.89512 sec, ips: 11.02570 instance/sec.
[11/26 12:27:46] epoch:[  1/5  ] train step:1    loss: 2.46962 lr: 0.001000 top1: 0.10000 top5: 0.50000 batch_cost: 0.00532 sec, reader_cost: 0.00132 sec, ips: 1880.51650 instance/sec.
[11/26 12:27:46] epoch:[  1/5  ] train step:2    loss: 2.26830 lr: 0.001000 top1: 0.10000 top5: 0.70000 batch_cost: 0.00277 sec, reader_cost: 0.00032 sec, ips: 3616.40283 instance/sec.
[11/26 12:27:46] epoch:[  1/5  ] train step:3    loss: 2.56208 lr: 0.001000 top1: 0.10000 top5: 0.60000 batch_cost: 0.00284 sec, reader_cost: 0.00029 sec, ips: 3522.55312 instance/sec.
[11/26 12:27:46] epoch:[  1/5  ] train step:4    loss: 2.78659 lr: 0.001000 top1: 0.00000 top5: 0.50000 batch_cost: 0.00277 sec, reader_cost: 0.00024 sec, ips: 3614.53292 instance/sec.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}