Allow hr to be a parameter in apply_hr

[pierucci]

[jrdnmdhl]

This presents a bit of a problem. Parameters might depend on survival distributions, but survival distributions may also depend on parameters. This would be a lot easier if survival distributions (and other arbitrary R-objects) could be parameters, but that would require a complete overhaul of parameter evaluation.

That overhaul might be worthwhile, though. It would make it much easier to do self-contained models (e.g. where the estimation occurs within the parameters table).