re-parameterization of Multi30K dataset

[parmeet]

Summary:

• Refactoring Multi30K translation dataset in accordance with re-parameterization of IWSLT dataset #1191
• Moved it to experimental folder
• Added example usage
• Added new doc page for experimental raw datasets
• Updated doc string

[codecov]

Codecov Report

Merging #1207 (d6e3a80) into master (27f9ed0) will increase coverage by 0.02%.
The diff coverage is 90.58%.

@@            Coverage Diff             @@
##           master    #1207      +/-   ##
==========================================
+ Coverage   73.21%   73.23%   +0.02%     
==========================================
  Files          67       67              
  Lines        3689     3718      +29     
==========================================
+ Hits         2701     2723      +22     
- Misses        988      995       +7     

Impacted Files	Coverage Δ
torchtext/datasets/__init__.py	100.00% <ø> (ø)
torchtext/experimental/datasets/translation.py	75.00% <80.00%> (-0.31%)	⬇️
torchtext/experimental/datasets/raw/multi30k.py	91.13% <91.13%> (ø)
torchtext/experimental/datasets/raw/__init__.py	100.00% <100.00%> (ø)
torchtext/data/datasets_utils.py	91.05% <0.00%> (-0.82%)	⬇️

---

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 27f9ed0...d6e3a80. Read the comment docs.

[cpuhrsch]

nit: the "in" operator should return a key by default, so I don't think you need .keys() here