Config files

[Jean-Baptiste-Camps]

Implement the idea of having config files, giving all the options for corpus creation, which could facilitate having several sets of features. It could be some sort of yaml, like:

sampling:
  sample_size: 1000
  sample_random: True
  max_samples: 1000
features:
  feature:
    type: words
    n: 1
    feat_list: functionwords.txt
  feature:
    type: affixes
    k: 500
  feature:
    type: pos
    n: 3
    k: 500

etc.