bgg.prepare.data: Prepare the dataset for the mining stuff
In 9thcirclegames/bgg-analysis: What makes a game a good game?

This function prepares the dataset removing unuseful character attributes, removing videogames (and videogames-only attributes).
\ These are the performed operations:

It removes the details.name attribute, as it's a redundand ambiguous identifier, and the stats.median column, as it seems to be broken in BGG XMLAPI2.
Some exploratory analysis bring up that lot of games have yearpublished=0: this is probably a missing set to zero by mistake, so we put to NA (we are aware that we have some very ancient games in the dataset, but we consider them as outliers).
Finally, we removed games with yearpublished < 1959 (.05 percentile and really unuseful).
We are aware that there are some games with yearpublished > 2015, but we decided to keep those yet unreleased games as they could be useful as test set.