sequence_mask fix: when the input length is an empty tensor, the kernel tries to dereference illegal sentinel iterator (#49525)

* add isort config

* isort all files

* update config

* re-blacken python code

* temporarily disable date and diff_py_file

* skip a format

* [CodeStyle][F401] remove unused imports in unittests/distributed_passes,tokenizer,sequence

* add noqa after required imports

* 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

API/OP(sequence_first_step, sequence_last_step, sequence_mask, beam_search, beam_search_decode) error message enhancement (#24590)

* move sequence op unittests to a separate folder
test=develop

* add missing CMakeList file
test=develop

* fix relative path import
test=develop

* fix relative import
test=develop

* use sys.path.append
test=develop

* sequnce mask support max length tensor input; test=develop

* add rnn_impl.py; test=develop

* add basic gru lstm unittest; test=develop

* fix api spec; test=develop

* fix sequence_mask op bug;
test=develop
test=document_preview

* change +-*x to elmentwise_op; test=develop

* add mkl flag; test=develop

* fix rnn impl bug; test=develop

* update api spec; test=develop

* fix doc bug; test=develop

* fix lstm bugs; test=develop