Store times as integer multiples of the master pseudoclock clock_resolution

[philipstarkey]

As per #99, floating point rounding errors can lead to incorrect compilation errors.

We should really avoid internal floating point storage of time as it just causes us grief. We'll probably need to store the floating point time for error messages (otherwise quantisation of times could lead to confusion when trying to identify the line of code causing the issue), but that's probably not any harder than the rest of the work needed.