* Polish cn doc fo gru_unit, lstm_unit and scaled_dot_product_attention.

* Add gather_tree_cn.rst and update the data type of assign/fill_constant_batch_size_like/gather_nd.

* Fix _cn_api_fluid_layers_gather_tree

* Update tensor_array_to_tensor and add mse_loss.

* update doc/fluid/api_cn/layers_cn/add_position_encoding_cn.rst
update doc/fluid/api_cn/layers_cn/beam_search_cn.rst
update doc/fluid/api_cn/layers_cn/beam_search_decode_cn.rst
update doc/fluid/api_cn/layers_cn/chunk_eval_cn.rst
update doc/fluid/api_cn/layers_cn/dynamic_gru_cn.rst
update doc/fluid/api_cn/layers_cn/gru_unit_cn.rst
update doc/fluid/api_cn/layers_cn/lstm_unit_cn.rst
update doc/fluid/api_cn/nets_cn/scaled_dot_product_attention_cn.rst

* Update format for dynamic_gru_cn, gru_unit_cn and scaled_dot_product_attention_cn.

* Fix problems of chinese docs

* Revise chinese docs

* Add more fixes

* Add data type comments for some inputs

* update conv2d 
conv3d
conv2d_transpose
conv3d_transpose
batch_norm
cos_sim
cross_entropy
npair_loss
dropout
img_conv_group cn doc

* fix api_cn of 11 APIs
test=develop
test=document_preview

* fix format of math and params

* fix tensor dimension description