[Cherry-pick] fix crf bug in paddlenlp 

* new chunk_eval

* np first

* fix readme and change cross-entropy api

* change api back

* add the benchmark glue for some task

* fix the qnil result for the bert benchmark

* change the from the logger to framework logger

* fix bleu bug

* update lrscheduler

* minor fix

* add pre-commit

* minor fix

* Add __len__ to squad dataset

* minor fix

* Add dureader robust prototype

* dataset implement

* minor fix

* fix var name

* add dureader-yesno train script and dataset

* add readme and fix md5sum

* integrete dureader datasets

* change var names: segment to mode, root to data_file

* minor fix

* update var name

* Fix api bugs

* add dataset readme

* add express ner

* update readme format

* fix format bug

* change readme path

* fix format bug

* fix dataset bug

* add 2 datasets and update squad metric

add ernie_gen for the dygraph mode

* add metric readme doc

* add "`" for APIs in doc

* Update README.md

Add "`" for APIs in doc

* Update metrics.md

* Update README.md

* update metric readme

* update rouge-l, move default_trans_func to utils

* fix error info bugs

* add score method for metric bleu and rouge-l

* update bleu doc

* update ppl, inherited from Metric, suitable for common mt and lm task.

* fix ppl bugs, delete useless argument

* fix ppl bugs, return a num not a numpy array.

* Update paddlenlp.metrics README

* Add info for Mcc

* delete DuReader metric

* update task name, add API col

* add electra pretrain and modify style of electra modeling

* add electra pretrain, modify style of electra modeling and fix problems of review

* delete predict_classifer

* modify accu to acc

* add paddlenlp.metrics.glue