Subset of annotations for FSD 1st release

[edufonseca]

After defining the mapping to create lists of candidate samples to fill the sound categories (#25), we should decide a subset of all mapped annotations where to concentrate the validations in the initial validation task for FSD 1st release. This is related to the design of the subsets #24, in such a way that the chosen subset of annotations should meet the requirements of the data subsets that we want to provide.

A simple example:
Suppose we want the Medium subset of the FSD 1st release to have ~100k annotations with rater agreement. When validating an annotation, raters can say NP or U with certain probability. This means that we should select >100k annotations as starting point.

[edufonseca]

We could do this in two ways:

1. assuming a constant factor for all categories, eg, if we want to have 10 (Present) audio clips per category, we select 15 clips (annotations) to be validated
2. considering that the mapping may be better in certain categories, eg, for a well mapped category, 13 audio clips may be enough, while for a bad mapped one we could need 20

[xavierfav]

The prioritization of categories according to their number of ground truth annotation should bring us to a balanced dataset #38.
We also choosed to omit the category that have two few annotation candidates.

I think this is enough for getting a good subset.
However, annotation are not prioritized based on sound "quality" #23

[xavierfav]

We decided not to create any subset, but rather prioritize some categories and annotations.
Some requirements are already functional (prioritize categories), and it will be improved in #133.