make hazard calculations threaded

[Kevin Milner]

Hazard curve calculations could easily be threaded, with different threads doing the calculation for different sources in an ERF. Local calculations could be spawned with a thread for each core, or a thread for each core - 1.

One issue is that the IMRs would have to be cloned or re-instantiated for each thread, so that the internal variables don't get overwritten. This shouldn't be a big issue though, as IMRs don't use much memory and instantiate very quickly.