[core] Slightly extends allocator interface
[ep] OpenCL EP
[op] Add basic elementwise op
[op] Add generic and depthwise conv2d for Conv
[op] Add Clip
[op] Add ReLU
[op] Add MaxPool for NCHW tensor
[op] Add GlobalAveragePool for NCHW tensor
[op] Add basic Resize (bilinear) implementation
[op][discuss] Add Shape for OpenCL EP
[core] Minor change graph_partitioner for opencl ort file format support
[op][Jicheng] Add Concat for opencl
[op][Jicheng] Fuse op Add and Relu
[op][Jicheng] Support nearest Resize
[op][Jicheng] Winograd support for Conv
[op][Jicheng] simplify winograd generator