Predictions on larger triangles

[jbogaardt]

When running predictions on triangles with more periods than are supported by the underlying model, we get silent failure of ultimate projections - ultimates are zeroed out. Offending code:

import chainladder as cl

model = cl.Chainladder().fit(cl.load_sample('ukmotor'))
new_tri = cl.load_sample('prism')['Paid'].sum().incr_to_cum().grain('OYDY')
model.predict(new_tri).ultimate_