1. improve n-fold quantification algorithm by introducing a minimal size for each fold. 2. automatically search for best n for n-fold algorithm. (#2167)

* 1. add quantification_fold parameter. 2. support quantification test in run.py.

* implement n-fold quantification

* open cl mem optimise, split with cpu codes.
fix a bug when some memory is not equal 4 .

* open cl mem optimise, split with cpu codes.   fix a bug when some memory is not equal 4  fix bad des

* get opencl kernel_work_group_size and set local_work_group_size for conv and conv_trans kernel of GAN, test=develop

* fix fpga CI error, test=develop

* fix fpga CI error, test=develop

* auto set local work group size according to devices

* Optimal OPENCL conv_trans_7x7 Performance , test=develop

* fix CI error, test=develop

* fix fpga ci build error. test=develop

* fix fpga ci build error. test=develop

* fix new_scale & new_bias crash test=develop

* fix new_bias & new_scale in GPU test=develop

* add opencl  depthwise_conv_trans_op

* test=develop

* Optimal OPENCL conv_3x3S2 Performance

* test=develop

* Optimal OPENCL conv_trans_3x3s2 Performance , test=develop

* add opencl conv_trans_op, test=develop

* add opencl  depthwise_conv_trans_op

* test=develop

* Optimal OPENCL conv_3x3S2 Performance

* test=develop

* Optimal OPENCL conv_trans_3x3s2 Performance , test=develop

* add opencl  depthwise_conv_trans_op

* test=develop

* Optimal OPENCL conv_3x3S2 Performance

* test=develop

* fix vector out of range crash test=develop

* fix style test=develop

* fix ios CPU preprocess test=develop

* fix style test=develop

* accelerate opencl compilation time test=develop

* fix style test=develop

* fix variant memory leak test=develop

* fix variant memory leak & fix style test=develop

* fix cpplint test=develop

* 1.add density_prior_box for gpu.
2.add flatten2 for gpu.
3.add concat 4 inputs size for gpu.
4.fix pool.
5.fix transpose2
test=develop

* 1.fix GetGPUTime
test=develop

* add opencl  depthwise_conv_trans_op

* test=develop