add minimum and maximum ranges for the predictions

[koaning]

There's an interesting language opportunity here.

LinearTrend(n_changepoints=10) * FourierSeasonality(n=5, period=12/143) * Range(0, 100)

vs.

LinearTrend(n_changepoints=10) + FourierSeasonality(n=5, period=12/143) * Range(0, 100)

In the former, the range effect is applied against everything. While in the latter the range is only applied to the seasonal effect. This is very sensible in many situations because you can often say "well the effect of season should only be this big".

Another reason why this is interesting from a grammar perspective is that it is easy to declare a type of prior belief. It might be reasonable though to allow for this too;

LinearTrend(n_changepoints=10) + FourierSeasonality(n=5, period=12/143) * Range(None, 100)

This could suggest that there is no lower limit of the seasonal effect, but there is an upper limit.