Optimizing the output of classification probablity in demo\text_classification\predict.py and corresponding README.md

Mismatching between data mode description and the corresponding code in demo\text_classification\README.md (#1762)

* Matching mistaken between demo\sequence_labeling\README.md and demo\sequence_labeling\train.py

* Matching mistaken between demo\text_classification\README.md and demo\text_classification\train.py

* Mismatching between data mode description and the corresponding code in demo\sequence_labeling\README.md

* Mismatching between data mode description and the corresponding code in demo\text_classification\README.md

* Add embedding seq-cls finetune demo and update api

* Update docs of pad_sequence and trunc_sequence

* Add electra modules

* Amend module versions of chinese-bert-wwm, chinese-bert-wwm-ext, rbt3 and rbtl3

* Updata demo README.md

* Add tutorial about nlp dataset customization

* Update doc and tokenizer usage in seq-cls demo

* Updata tutorial

* Add MSRA dataset

* Add ernie_tiny seq-cls task case

* Update README.md

* update docs

* update docs

* add finetune to module tutorial; add 1.6 release note

* add bertmodule

* update doc

* update docs

* update docs

* update paddlehub_finetune.gif

* update docs