Adding ftfy to fix encoding issues

[albarrentine]

Hey all,

As mentioned in openaddresses/openaddresses#2392, this PR adds the supremely handy ftfy to machine for fixing common encoding issues like Mojibake, etc. Should also help in cases where there's mixed encoding, say if a source contains some legacy data with a different encoding that was never converted after a migration.

As a dependency it's a simple pure-Python library so doesn't add much weight, works on Python 2 & 3, etc. I use it whenever I have to deal with text from the web, including on the Python/ingestion side of libpostal when importing data sets like OpenAddresses and OSM.

It works on Unicode, so essentially just wrap every call to .decode with a call to ftfy.fix_encoding. Here's a (Python2) usage example:

import ftfy
input = u"Брестская обл., Столинский р-н, Белоушский с/с, д. Рыбники, ул. Толстого, д. 17"
ftfy.fix_encoding(input)
# returns: "Брестская обл., Столинский р-н, Белоушский с/с, д. Рыбники, ул. Толстого, д. 17"

At present my dev setup for machine is not fully baked so using the PR to run tests.

[migurski]

Looking good, I’ll keep an eye on this PR!

[albarrentine]

Ran tests locally but saw the same test fail on master as on this branch. Is that a known issue?

[migurski]

The failing test was added here, to uncover GDAL 2.x related issues: #514

It looks like ftfy is returning with a null value?