Restart dataloader - Support for deterministically going through the data loader upon restart

[lorr1]

Description of the problems or issues

There is currently no flag to iterate over data loader if you want to restart your model at the same spot in the data.

Description of the proposed changes

If user provides steps_learned and has skip_learned_data turned on, the I set the start_step to be 0 and the train step to be steps learn. I increment the batches without calling forward of the model until I reach train steps.

Test plan

Made the argument test pass. I have manually tested the resume behavior and that the model does continue when learned steps are set.

Checklist

• [ ] I have updated the documentation accordingly.
• [ ] I have added tests to cover my changes.
• [ ] All new and existing tests passed.
• [ ] I have updated the CHANGELOG.rst accordingly.

[codecov[bot]]

Codecov Report

Merging #101 (e5068a1) into master (3475115) will decrease coverage by 0.31%. The diff coverage is 28.57%.

@@            Coverage Diff             @@
##           master     #101      +/-   ##
==========================================
- Coverage   85.31%   85.00%   -0.32%     
==========================================
  Files          40       40              
  Lines        1955     1967      +12     
  Branches      418      422       +4     
==========================================
+ Hits         1668     1672       +4     
- Misses        171      177       +6     
- Partials      116      118       +2     

Flag	Coverage Δ
unittests	85.00% <28.57%> (-0.32%)	:arrow_down:

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
src/emmental/learner.py	67.51% <23.07%> (-1.47%)	:arrow_down:
src/emmental/utils/parse_args.py	99.32% <100.00%> (+<0.01%)	:arrow_up: