The influence of centre selection on the generalisability of economic evaluations conducted alongside randomised controlled trials: a case study from the Rossini trial