Prioritize annotations corresponding to short clips about freesound-datasets HOT 2 CLOSED

We can think of what happens when, in a category, sounds <10s are already validated.
Currently files are in the range [0:90] seconds, with 40% of the samples <6s.
Sounds' variable length implies the need for certain classification approaches.

When it comes to sounds' duration, for FSD 1.0 we could understand “short” files as the most interesting (and abundant) and try to avoid large differences in duration. As suggested in ISMIR 2017, we could prioritize the annotation of short sounds according to the following ranges:

1. 0-10s (already implemented)
2. when 1) is over: 10-20s
3. when 2) is over: 20-30s

Goal: a more homogeneous release in terms of sound duration (probably easier to handle and more reasonable)

from freesound-datasets.

This issue will be addressed in #133.

from freesound-datasets.