Speed up combining dataframes

In [these lines](https://github.com/bcgov/epi.branch.sim/blob/69c2fd9f422cdcea6caa06d29facc74a921ba5ca/R/step_simulation.R#L158) of `step_simulation()` we use `rbind` to combine the data frame of existing cases with the data frame for new cases. This happens at least once in almost every time step and is potentially a major bottleneck in terms of speed.

Pre-allocating a large data frame, matrix or list would speed things up, however, that would require some adjustments throughout the package as the `state_df` data frame has rows removed as well as added. This requires a bit more effort and will be a goal for the next major release.

Start in R/step_simulation.R at the step_simulation() code around the linked rbind call. Trace how state_df rows are added and removed across the package, then identify the affected benchmarks or simulation checks. Done means reducing the repeated combining cost while preserving the existing simulation behavior.

Written by the indexing model from the issue text.