* bn pattern

* 1. add conv bn fusion pass(conv2d, depthwise_conv2d)
2. modify the graph to ssagraph
test=develop

* simplify fuse process.

* update from upstream

* close the kernels not empty checker.

* add the pattern matcher logic. test=develop

* update op_info usage. test=develop

* fix conflicts

* fix kernel registry realted bugs test=develop

* add fc unit test

* refine eigen fc
add cpu info, arm context
init packed sgemm

* enable packed sgemm

* add arm math

* pass fc ut

* follow comments

* add cmake

* update

* fix proto pd

* fix compile

* tmp save

* fix protobuf device version

* fix protobuf and host compile

* fix std c++11 support on android

* change array to vector to fix ndk c++_static

* fix rt and add dockerfile

* fix android compile issue with latest merge

* init arm kernels

* enable run on arm

* update format

* update format

* update format

test=develop

test=develop

test=develop

make typesystem simpler

Instruction to Stmt

to support mobile

by default

and code clean

MUL CUDA kernel is faked, for the cublas not works

Some enhancement needs for TypeSystem, unittests

- make the target wrapper for host works
- code clean