Empirical contact distribution

One aspects that has arisen when discussing the difficulty of simulating realistic line list and contacts data is parameterising the contact distribution (`contact_distribution` argument). This is rarely, if ever, reported in the epidemiological literature and instead data such as contact surveys or contact matrices are more commonly reported. Therefore to make it easier to emulate empirical contact patterns it may be useful to allow user to pass an empirical contact distribution (`numeric` vector) to the `contact_distribution`.

One example dataset that could be a good test of this feature is the contact distribution from [Kucharski et al.](https://www.thelancet.com/journals/laninf/article/PIIS1473-3099(20)30457-6/fulltext?ref=witsio), with the [data shared on GitHub](https://github.com/adamkucharski/2020-cov-tracing/blob/master/data/contact_distributions_o18.csv).

An initial sketch of how to implement this feature is on the `empirical-contact-dist` branch.