Calculation of .full_expectation in CaseOutstanding()

[kennethshsu]

The first issue in #159, there is currently no research literature on how we can estimate residuals for the CaseOutstanding method. As such, the method currently codes the upper triangle with known values.

Test code:

import chainladder as cl
qtr = cl.load_sample('quarterly')
qtr = cl.CaseOutstanding(paid_to_incurred=('paid', 'incurred')).fit_transform(qtr)
cl.Chainladder().fit(qtr).full_expectation_['paid'] - qtr['paid']

For the full discussion, see the original post by @jbogaardt in https://github.com/casact/chainladder-python/issues/159#issuecomment-1069049003