Replace `assert np.close` with `np.testing.assert_allclose` in collective communication unittests (#49195)

* refactor: replace `assert` with `assert_allclose`

* chore: add coverage conf

* revert: remove incorrect coverage conf

* generate static graph code of some ops

* change the default value of 'num' of 'unstack'

* revert the pow

* fix the 'real' 'imag' op error because of 'complex'

* fix the code according to review

This reverts commit 9df0ab32.

* fix onednn squeeze bug

* add test

* update kernel

* remove unittests eager guard tests part2

* rm bicubic eager guard tests

* review

* fix

* fix dtype bug in bce_loss_with_logits

* rm eager guard tests part3_2

* modify

* gn nhwc8

* remove error

* fix api docs format problems 121-131

* fix English docs #49042

* resolve conflict; test=document_fix
Co-authored-by: Ligoml <limengliu@tiaozhan.com>

* delete FLAGS_enable_eager_mode

* move dropout_impl from fluid to phi

* move cuda_graph_with_memory_pool from fluid to phi

* update namespace

* remove cuad_graph in fluid

* fix mac-build

* fix bugs

* correct CodeStyle

* fix mac-build

* fix mutable_data

* fix stl include

* fix copy param

* move gather_scatter_kernel from fluid to phi

* mv gather_scatter_kernel to gather_scatter_functor

* solve conflicts of remove_lod_reset and latest develop version

* add lod_reset to the no_grad_set_white_list.py

* 移除 declarative 接口

* 修改 CI报错，清理注释

* remove fluid.mul

* remove mul in __all__

* rm unittest eager guard part1

* review

* reset spwan runner base

* reset op_test

* rm test_type_core

* rm imperative op eager guard tests

* modify varbase patch eager guard doc

* restart looup_table_v2

* General optimization for no_varlen embedding layernorm

* add 0-d support for paddle.kthvalue

* add 0-d support for paddle.mode

* fix coverage test for device