Add XLM-RoBERTa model for TF2

[jplu]

Hello,

I have implemented the XLM-RoBERTa model handling for Tensorflow 2.

[jplu]

There is a little incompatibility between isort and black apparently psf/black#251

[codecov-io]

Codecov Report

Merging #2454 into master will increase coverage by 0.6%.
The diff coverage is 100%.

@@            Coverage Diff            @@
##           master    #2454     +/-   ##
=========================================
+ Coverage   74.51%   75.11%   +0.6%     
=========================================
  Files          87       88      +1     
  Lines       14920    14945     +25     
=========================================
+ Hits        11117    11226    +109     
+ Misses       3803     3719     -84

Impacted Files	Coverage Δ
src/transformers/__init__.py	98.8% <100%> (+0.01%)	⬆️
src/transformers/modeling_tf_xlm_roberta.py	100% <100%> (ø)
src/transformers/data/processors/squad.py	56.72% <0%> (+27.54%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9d87eaf...bb1aa06. Read the comment docs.

[houdaM97]

does it work even if xlm-roberta-large is pretrained pytorch model? i mean do we need to convert pytorch model to tensorflow?

[LysandreJik]

@jplu I took the liberty of updating the documentation to the new format directly on your fork. Thank you for your contribution, this is awesome!