update tensorrt to 8.2 on aarch64

[sanblch]

These changes make possible to run apollo on Jetson Xavier AGX with Jetson r32.6.

1. Use of tensorrt api updated according to changes at tensorrt 8.2. Biggest changes are switch IPlugin -> IPluginV2, DimsNCHW -> Dims4, DimsCHW -> Dims3 See 1, 2.
2. Little arch conditionals addition to make it compile on aarch64 like remove SSE opt, add #ifdef __aarch64__, and include to build inside conditional if_x86_64().

[CLAassistant]

All committers have signed the CLA.

[ycdhqzhiai]

@sanblch I run darkSCNN according to the code you modified, but all the output is 0. Have you ever met this situation？

[QiTianDaShengDaShi]

Have you solved this problem?

[QiTianDaShengDaShi]

@sanblch我按照你修改的代码运行darkSCNN，但是输出全部为0，你遇到过这种情况吗？

请问你解决这个问题了吗，我叶遇到了

[ycdhqzhiai]

@QiTianDaShengMaoHuiFei 没有，我还是用的7

[jumogehn]

@sanblch Can I use this patch for x86_64 also?

[cherishTMYY]

Jetson Xavier AGX

@QiTianDaShengMaoHuiFei 没有，我还是用的7

请问您是在Jetson Xavier AGX上跑的tensorRT_7吗？能成功吗？