Run without population file

[galactic-src]

Currently it is difficult to configure a quick, basic simulation. This sort of thing would be useful for creating regression tests that target particular features.

This PR involves a few changes and probably will need some tidying:

• require that without a population file you supply the simulation grid width and height (rather than inferring it from the population size, which ended up with a very sparse grid, depending on number of microcells per cell)
• set height and width of cells (P.in_cells_) immediately after reading the parameter, or inferring the default
• refactor population allocation loop to incorporate the last iteration. (I would love to rename some variables here, but thought it might be easier to follow without including in this PR)
• only validate adunit values if adunits are being included in the simulation

These changes allowed me to run a simulation from some really stripped down parameter files, albeit without any means (I think) for infection to spread across cell boundaries. I am also limited to a low infectious period by the MAX_INFECTIOUS_STEPS validation (different validation according to whether P.DoInfectiousnessProfile is set, which I am avoiding for a minimal config).

Keen for some feedback - am I on the right lines?

A possible fix for https://github.com/mrc-ide/covid-sim/issues/337