[cherry-pick] Fix build error caused by flatbuffer if the target is tiny_publish, and fix the missing of the attr name of the output scale (#4356)

* [LITE][XPU] 1. Add sequence_unpad kernel for XPU; 2. Bugfix in sequence_unpad kernel for x86, as InferShapeImpl() is now empty in lite/operators/sequence_unpad_op.cc; 3. Refine TargetWrapperXPU; (#4237)

* [Core] Fix the missing of the attr name of the output scale (#4334)

* [NPU] Fix build error caused by flatbuffer if the target is tiny_publish (#4340)
test=develop
Co-authored-by: NCwndmiao <miaotianxiang@baidu.com>

[RKNPU] Adapting to the changing of the quantization parameters, and supporting the fully quantized mobilenetv1 model (#4046)

* [RKNPU] Adapting to the changing of the quantization parameters, and supporting the fully quantized mobilenetv1 model
test=develop

* [Doc] Update the doc for apu and rknpu
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

* [NPU] apply npu cache offline model to other devices, test=develop

* [NPU] address review comments, test=develop

* reduce .so size. test=develop

* compile all targets when LITE_ON_TINY_PUBLISH=OFF

* unordered_map is more convenient when key is customized class

* test=develop