Add a simulation mode for consideration sets. While fiddling with modes, remove the typed mode (for now), which wasn't used much. This considerably cleans up internal code, as typed games were a special case. Also fix a potent miscalculation in the weighted best-reply function, which wrongly accumulated strategy-switching probabilities. Lastly, fix the randomised starting point to properly randomise over all strategies and players: there was unintentional bias in the division of strategies into agent populations. Update the Operator's Guide to reflect the algorithm used for this (in terms of bias).

In terms of usability, considerably update the documentation to clarify the concept of a simulation mode, which combines selection, reproduction, and a death process. To wit, make the Moran death process mandatory for some modes (payoff reproduction and mutation, etc.) as I can't conceive of not using it. Note this in the documentation. This is all reflected in the shortened and polished Operator's Guide. While there, make the Operator's Guide look a bit better. Finally, add a mean view to the non-ergodic view's population share analysis.