[LITE][OPENCL] Fix opencl backend: Free opencl resources; Fix AddKernel/GetKernel, program and all opencl kernels (#3344)

* [DONT MERGE] Fix opencl backend.

* [LITE][OPENCL] Fix kernels overlapped when add/get for kernels of mnasnet/yolonano. test=develop

* remove useless. test=develop

* add all image kernels for Get/Add kernel. test=develop

* add all image kernels for Get/Add kernel. test=develop

* fix buffer kernels of opencl. test=develop

* fix release opencl. test=develop

* [LITE][OPENCL] improve perf and support variable length for fc_buffer、elementwise_add、scale、activation、grid_sampler. test=develop

* [LITE][OPENCL][Image] add lws turn &  close cl check when shutdownlog , test=develop

* [LITE][OPENCL][Image] add lws turn &  close cl check when shutdownlog , test=develop

* [LITE][OPENCL][Image] add lws turn &  close cl check when shutdownlog , test=develop

* [LITE][OPENCL][Image] add lws turn &  close cl check when shutdownlog , test=develop

* [LITE][OPENCL][Image] add lws turn &  close cl check when shutdownlog , test=develop

* split precision profiler from performance profiler. test=develop

* add conv_transpose+bn fusion. test=develop

* delete note, test=develop

* fix format space, test=develop

* fix opt run error, test=develop

* [LITE][BM] add fill_const assign_value ops, test=develop

* add train_demo, test=develop

add cuda kernel.

abs, tanh, elementwise_sub

* fix format, test=develop

* add some op infershape implement, test=develop

 fix sequence_pool bug

* [LITE][OPENCL] Add ImageFree for Free in memory. test=develop

* add initialize for cl_use_image2d_ when use buffer. test=develop

* [LITE][OPENCL] fix elemul kernel of opencl. test=develop

* fix instanceNorm of opencl. test=develop

* add more info about cl device info. test=develop