No longer depends on python-Levenstein

(But if installed, will use it for a speedup)

Fix divisionbyzero error

v0.0.4

Update setup.py

Improved matching performance for big datasets

Many changes:

• Probabalistic matching now on a field-by-field basis rather than lumping all tokens into one
• Search algorithm can now account for scores of matches found, stopping when a good match is detected
• Better scoring algorithm that accounts for tokens which don't match and misspellings
• Memoisation to improve performance

v0.0.1

Basic functionality working