while filter weight is quanted elementwise, use maximum of
weight_scale instead of mininum to keep quant weight in range of int8

           testcase

[MLU] add some basic support for MLU, including related passes, kernels, gtests and some api in padddle_api.h
Passes：mlu_subgraph_pass ,mlu_postprocess_pass
Kernels:  act，batch_norm, concat, conv, elementwise, fc, interpolate, pool, scale, softmax

destroy base op after fuse

(ref): add sigmoid and elementwise_mul x86 kernel interface

    1. disable conv activation pass by default
    2. set fc_fuser'param with_relu false while mlu fc kernel does not
    support relu
    3. change fc filter shape from 2 dim to 4 dim while input dim == 4
    4. add ToFile func in mlu tensor for debug convenience
    5. enable 4-dim input in elementwise_ops
    6. add transpose2d in utility.cc

target -> subgraph -> target -> precision -> layout

add concat and scale to paddle_use_bridges.h