Make sure we load correct data

[willdurand]

This PR makes the SGD/CGD data import slightly more robust:

• it ignores too long aliases (> 100 chars)
• it skips and logs rows with no name

[jmaupetit]

Great! Just a single logging statement to add.

[jmaupetit]

I would add a warning message when we ignore one or more aliases:

splitted_aliases = feature['aliases'].split('|')

aliases = filter(
    lambda a: len(a) <= 100,
    map(str, splitted_aliases)
)

if len(aliases) != len(splitted_aliases):
    logger.warning('Ignored long aliases (100+ chars) in {}'.format(feature['aliases'])

[willdurand]

good call

[jmaupetit]

👍

[coveralls]

Coverage decreased (-0.01%) to 99.366% when pulling 4fcf8b6 on long-aliases into 2d52a31 on master.