* replace paddle.fluid.layers.data and remove io.data

* partial commit

* partial commit

* partial commit

* partial commit

* partial commit

* partial commit

* remove data in fluid.layers.io.__all__

* fix errors

* fix unitests

* fix unitest

* fix unitests

* fix unitest

* fix unitest

* fix unitests

* fix unitest

* fix test_layers unitests

* fix typro

* fix unitest

* fix unitest

* fix unitest

* fix typro

* fix unitest test_model_cast_to_bf16

* fix test_reducescatter

* fix collective unitest

* fix collective unitests

* fix collective unitests

* add coverage

* fix add layers.data

* re run ci

* fix some typro

* fix samplecode error

* fix samplecode error

* move fc from fluid to paddle2.0

* fix unit test

* fix some examples

* fix some examples

* modify slice infershape

* code style

* modify slice_unittest

* temp fix

* batch_norm api move

* code_style

* codestyle

* ci_static

* add __init__

* reset other change

* revert .cc

* add import batchnorm

* conflict and revert

* fix bug

* fix third conflict one day

* fix conflict

* fix conflict bug

* fix conflict bug

* modify api

* code_style

* modify doc

* add lost doc stable

* fix conflict bug

* ci lack of gpu

* add isort config

* isort all files

* remove apis in fluid.ops

* fix test_activation_nn_grad

* fix circle import error

* fix ops

* fix cos

* fix divide not inplace

* remove lazy-import part

* update config

* re-blacken python code

* temporarily disable date and diff_py_file

* skip a format

* change fluid.mean to paddle.mean

* reverse some old code examples

* 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

Add tests to use dy/dx to make sure the gradient values calculated by the control flow backward is correct. Also fixed bugs detected by those tests.

Fix bugs:

1. Unlike sum_op, optimizer ops don't allow uninitialized input tensor. But in conditional_block_grad_op, since the conditional_block may not run, the output gradient tensor may be uninitialized, which will cause the optimizer op error. To fix it, we should let optimizer ops support uninitialized input like sum_op or assign the uninitialized gradient to 0 when the conditional_block_grad_op doesn't run. I found there are about 10+ optimizer ops. **To be simpler, I just assign output gradient of the conditional_block_grad_op to 0 in this PR**. But it can be further explored whether we can make optimizer ops like sum_op to support uninitialized input tensor because theoretically we can speed up without the assigning in conditional_block_grad_op.

2. Infer parameter shapes during append_backward. I didn't know that all our parameters are in global block. When op_desc is inferring shapes at the sub-block, it may not know the shape of gradients of parameters whose shape information is at global block. I fixed it by inferring shapes of gradients from forward var.

This PR also did some code clean up:
1. Print the var name when sgd_op catches shape error so that it is easier to debug
2. Fix a typo: dicta -> dict

* unify DataLoader APIs, test=develop

* integrate iterable CPU Dataset, test=develop
add GPU dataset supporting, test=develop

* add unittests for dataset, test=develop

* add more docs to dataloader apis, test=develop, test=document_preview

* refine doc, test=develop

* refine doc again, test=develop

* increase coverage, test=develop

* support sparse gradients
test=develop

* add unit test for fuse_opt_ops
test=develop