Add more flexible aggregation in Trial to accommodate opinion dynamics next, or other social and cognitive sub-models

[mt-digital]

currently we only have summarise_prevalence, so we don't even have summaries of fitness, which will be important for reinforcement learning.

Tasks:

• as a first pass (v0.1.1, e.g.) write summarise_fitness to identify the common data flow
• it will be done when there is a summarise_series function to aggregate time series over many trials for a given parameter setting, generalizing the current summarise_prevalence