Enable to get the precise gpu_time for tf by using ctypes package to load libcudart.so and add add nchw support for batch_norm.

[Xreki]

该PR主要包括以下工作：

1. 添加控制benchmark行为的BENCHMARK_NEED_FEED和BENCHMARK_NEED_FETCH两个环境变量，并且在BENCHMARK_NEED_FETCH开启时，解析nvprof log时会减掉DtoH的时间

2. 更新tf支持的版本到最新版本2.7.0

[paddle-bot-old[bot]]

Thanks for your contribution! Please wait for the result of CI firstly.

[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.}