Text this: Re-analysis of data from cluster randomised trials to explore the impact of model choice on estimates of odds ratios: study protocol