* fix sequence pool grad compute error, test=develop

* fix sequence pool grad run, test=develop

* fix format test=develop

* add support for three-dimentional input in reduce max test=develop

* change pooling3x3s2max padding right to intrisic. test=develop

* fix arm winograd compute segment. test=develop

* add elu act

* fix elu act not find error, test=develop

* fix format.test=develop

* fix int8 model opt error in conv+conv fusion, test=develop

* fix format. test=develop

* test=develop

* [Core][ARM] Fix beam_search, eltwise_mul supports broadcast and int64_t data type, add print op and kernel, add exeception
test=develop

* Fix the dims of parent idx of the arm kernel of beam_search op

* elementwise_mul supports int64_t data type with broadcasting

* Add print op and kernel for debugging

* Support throwing the exception when the internal error occurs

* Refine while and conditional_block op kernel

* Support the graph optimization on subblocks

* Pass program_desc and block_idx into the kernel of the control flow ops(while/conditional_block/subgraph), and create the RuntimeProgram online, it make it possiable to call the control flow ops recursively

*Add unit test for masked transformer model

* add conv+conv(1x1s1p0) fusion

* fix build and run error

* fix formmat. test=develop

* update sequence_conv profile

* delete extra file, test=develop

* update sequence_pool and sequence_conv profiler, test=develop

* fix compute error, test=develop

* delete warning and extra info, test=develop

add act grad ops

* fix ttfnet miscs. test=develop

* fix conv int8 kernel choose and sooftmax compute bug

* change axis_size = 4 kernel choose, test=develop

* fix format. test=develop