[ENH] tSNE: Output preprocessor

[VesnaT]

Issue

Implements third part of #3342

Description of changes

Includes

• Code changes
• Tests
• Documentation

[codecov]

Codecov Report

Merging #3407 into master will decrease coverage by <.01%.
The diff coverage is 98.61%.

@@            Coverage Diff             @@
##           master    #3407      +/-   ##
==========================================
- Coverage   82.33%   82.32%   -0.01%     
==========================================
  Files         360      360              
  Lines       64045    64075      +30     
==========================================
+ Hits        52729    52751      +22     
- Misses      11316    11324       +8

[pavlin-policar]

I don't like list(range(n)). While it would very rarely come up, it would imply that the variables are named t-SNE-0, t-SNE-1, t-SNE-2, ... I guess it is consistent with the current behaviour of Manifold learning, which outputs C0, C1. Honestly, I think t-SNE-x and t-SNE-y should be removed in favour of t-SNE-1, t-SNE-2, and if the axes are to be labelled with x and y, this should be done on the t-SNE widget level.