* optimize preparation overhead before executing cinn compiled program

* update code notes

* fix flag annotation

* enable auto-tune when using CINN

* update cinn commit tag

* skip test

* fix lacking header file

[Phi]Separate AddKernel/DivideKernel/SubtractKernel/MultiplyKernel from ElementwiseKernel（Part1） (#41806)

* seperate add/div/sub/mul from elementwise

* delete code

* fix compile bugs

* deal with conflict

* fix bugs when compile

* fix windows unit test bug

* fix ci converage bugs

cinn_launch_op: optimize the overhead of preparing variables before executing cinn compiled program (#41777)

* optimize preparation overhead before executing cinn compiled program

* update code notes

* fix flag annotation

* add a flag of auto-tune feature beforehand

* remove needless deps in unittests

* add gpu marco

* fix other unittests

* fix kernel name error

* fix test_prepare_op

* fix failed dygraph unittests

* fix gpu failed tests

* fix cinn test failed

* fix cinn test failed

* fix dropout tests

* swith to PE execution in cinn launch

* fix outer variables erased

* skip the map bug temporarily for test

* temporary solution for batch_norm bug

* update comment

* fix compile error

* cinn_instruction_run_op_test: update code to skip external alloc/free instructions generated

* rename pten dir to phi

* rename namespace to phi

* rename infrt pten dir to phi

* resolve conflict

* rename pten to phi in cmake

* revert all infrt change

* change needed files

* fix infrt failed

* fix inference failed

* Unify paddle/pten::framework::ddim into pten::ddim

* fix paddle namespace

* compile sucessfully

* fix npu src file

* fix conflict

* fix conflict

* fix tensorrt compiler error

* fix conflict

* fix conflict

* fix tesst file conflict

* fix conflict

* fix mlu file conflict

* fix mlu file conflict

* fix cinn header file conflict

* fix conflict

* fix conflict

* fix conflict

* fix conflict

* add cinn_instruction_run_op test code

* update several interfaces of CinnLaunchContext

* update several interfaces and add detail comments in CinnLaunchContext class

* to skip the bug of error message check

* fix ut test failed due to reliant interface updated

* remove reshape and elementwise_add registry

* delete code

* fix bugs when run ci ut

* remove log

* fix bugs when run unit test

* fix bugs when run unit test

* fix bugs when run cinn

* fix bugs when run ci-mac-python3

* fix compile bugs

* fix compile bugs

* fix compile bugs

* fix bugs when run kunlun

* fix bugs when compile

* update code according comment

* Merged LoDTensor with Tensor,test=allcases

* Patched python level LoDTensor

* Patched python level LoDTensor

* Merge Tensor into DenseTensor

* Fixed namespace issues,test=allcases

* Fixed merge issues

* Fixed inference issues

* Fixed NPU test issues

* Fixed merge issues

cinn_launch_op： cache scope and place on CinnLaunchContext to skip duplicate alloc/free callback construction

1. add a subdirectory named `cinn` in `paddle/fluid/operators` directory and move releated files into it
2. seperate CinnLaunchContext class from `cinn_launch_op.h` and put it in a  new independent file named `cinn_launch_context.h`, so that it can be included by others clearly.

Modify ShareTensorWithCinnBuffer by callback to save memory

* optimize cache-key by replace GraphToProgram to Dot string

* fix compile failure bug

Modify serveral implements on CinnLaunchOp：
1. Skip checking input variables must be used 
2. Move current helper functions to a CinnlaunchContext 

cinn_launch_op: remove the check on extracting temporary variables

1. 功能不变，简化CinnLaunchOpKernel实现：将原先直接从Scope获取变量信息的方式改为借助参数ExecutionContext标准接口获取，简化了实现逻辑，相应地也简化了辅助函数的实现，原先cinn_launch_op_helper较为冗余，删除不必要的接口并迁移至cinn_launch_op.cc中定义。
2. 修复CinnLaunchOp InferShape判断是否有指定输出：HasOutput->HasOutputs
3. 添加详细的注释和debug信息，方便问题排查和代码维护

增加CinnLaunchOp，负责执行Cinn子图编译的结果，要点如下：
1. 在子图划分的BuildCinnPass中，每个子图在原图中会被替换为该CinnLaunchOp，由它来调用Cinn进行子图编译、执行的功能。
2. CinnLaunchOp的输入/输出即为子图的输入和输出，另外增加`compilation_key`属性，它可由该属性key从全局Cache中获取子图对象、编译结果，该属性由BuildCinnPass在创建Op时进行设置
3. CinnLaunchOp功能实现的流程为：
        - 从全局Cache中获取子图对象
        - 从全局Cache中获取子图编译结果，未命中cache时进行即时编译
        - 根据编译结果的变量信息(数据类型、shape）初始化运行时数据，分配内存/显存
        - 将运行时数据打包为参数，调用cinn的可执行对象runtime program进行计算
        - 子图运行结果通过参数指针同步到paddle侧的tensor