Uncertainty based on scenarios

[romainsacchi]

@Loisel
Instead of producing one database per scenario, we could build one database, in which inputs values would be defined as uncertainty ranges based on worst to best REMIND scenario variables values.

Before building the database, one could weight the likelihood of the worst and best scenarios of becoming true ("I believe that there is 60% chance that the world goes to shit, 35% chance that we will follow a moderate path and 5% chance we comply with the Paris agreement targets", which will probably look like a log-normal or beta distribution). This could help us shape as distribution probability that we could apply to technology market shares, efficiencies, etc.