[FEATURE] Add raw attention scores to the AttentionCell #951

[emilmont]

Description

For implementing a pointer mechanism in sequence to sequence models it is very practical to re-use attention cells. For example see the Attention-Based Copy Mechanism described in Jia, Robin, and Percy Liang. "Data recombination for neural semantic parsing." arXiv preprint arXiv:1606.03622 (2016).
The proposal is to additionally return the raw attention scores in the AttentionCell.

Checklist

Essentials

• PR's title starts with a category (e.g. [BUGFIX], [MODEL], [TUTORIAL], [FEATURE], [DOC], etc)
• Changes are complete (i.e. I finished coding on this PR)
• All changes have test coverage
• Code is well-documented

Comments

• The change is a backward compatible

cc @dmlc/gluon-nlp-team

[codecov]

Codecov Report

Merging #964 into master will increase coverage by 0.03%.
The diff coverage is 87.87%.

@@            Coverage Diff            @@
##           master    #964      +/-   ##
=========================================
+ Coverage   89.87%   89.9%   +0.03%     
=========================================
  Files          67      67              
  Lines        6371    6390      +19     
=========================================
+ Hits         5726    5745      +19     
  Misses        645     645

Impacted Files	Coverage Δ
src/gluonnlp/model/attention_cell.py	95.06% <87.87%> (+0.65%)	⬆️

[mli]

Job PR-964/1 is complete.
Docs are uploaded to http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR-964/1/index.html

[mli]

Job PR-964/2 is complete.
Docs are uploaded to http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR-964/2/index.html

[mli]

Job PR-964/3 is complete.
Docs are uploaded to http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR-964/3/index.html

[mli]

Job PR-964/5 is complete.
Docs are uploaded to http://gluon-nlp-staging.s3-accelerate.dualstack.amazonaws.com/PR-964/5/index.html