* remove layers.matmul in nets.py

* remove layers.matmul in rnn_impl/test_quantization_pass/auto_parallel_gpt_model/test_auto_parallel_completion_gpt

* remove layers.matmul in other files

* fix

* fix

* remove layers.matmul itself

* remove ref in CMakeLists.txt and tools directory

* remove matmul in fluid.layers.nn.py

* remove matmul in fluid.dygraph.rnn.py && resotre test_matmul_op.py

* replace matmul in fluid.dygraph.rnn.py && clean api_test in test_matmul_op.py

* fix error && restore empty test_auto_search_dist_matmul_op.py

* fix check in test_auto_parallel_partitioner.py

* fix test_dist_matmul && test_flags_mkldnn_ops_on_off

* fix test_fused_attention_op_xpu.py && test_matmul_op_xpu.py

* remove test_auto_search_dist_matmul_op.py

* remove layers.matmul in auto_parallel_gpt_model.py && fix doc in fluid/io.py

* fix for matmul_grad

* fix codestyle

* fix codestyle

* resolve conflicts error

* restore unit test file but not compiled it for later remove

* fix codestyle

* fix wrong unittest skip

* fix unittest delete

* fix scale cost

* fix scale cost

* resolve conflicts error

* resolve conflicts error
Co-authored-by: Njakpiase <jakpia21@gmail.com>

* isort all files

* revert conflicting files

* revert conflicting files

* revert conflicting files

* revert conflicting files

* revert conflicting files

* update config

* re-blacken python code

* temporarily disable date and diff_py_file

* skip a format

* fused_gate_attention manual code in eager

* use yapf to format all python file

* yapf exclude two unittests file for they rely on writing and reading file, and format will break them

* disable diff_py_file because too many diff files cause command following failed

* replace dropout_is_test with is_test.
* improve atol on a100.

Add support for bias is none for fused_attention op.

* fix bug:
1. atten: set the default value of attn_dropout_rate to None
2. ffn: add activation parameter

att, bug fix

目前的fused_attention_op不支持attn_mask=None的输入，本PR对此进行了补充，并补充了相应的单测逻辑。

* Fix bug when pre_layer_norm is false.

功能：本PR的目标是提高attention模块的计算性能。
为了减少框架层对op的调度开销，本PR通过在C++层手动实现attention模块，对外提供attention 大op；
为了减少防存开销，本PR采取了两种优化方法：
（1）在q,k,v计算时通过共享输入X，将该处的gemm，transpose和bias add从三次调用减少为一次；
（2）使用kernel融合优化技术，在不同cuda kernel之间通过寄存器传输数据；

将 #35905 和 #35843 PR中新增的的python api接口移到incubate目录下。

功能：本PR的目标是提高attention模块的计算性能。
为了减少框架层对op的调度开销，本PR通过在C++层手动实现attention模块，对外提供attention 大op；
为了减少防存开销，本PR采取了两种优化方法：
（1）在q,k,v计算时通过共享输入X，将该处的gemm，transpose和bias add从三次调用减少为一次；
（2）使用kernel融合优化技术，在不同cuda kernel之间通过寄存器传输数据；