Create "problem-downloader" helper

[tkornuta-ibm]

The idea of this helper is quite simple:

• load the problem
• check if data associated with given dataset exists
• if not, download/unpack/preprocess the data.

arguments: --p problem_classname (mandatory) --d data_dir (DEFAULT: ~/data/lowercase(problem_classname)