Move parallel for-loop into objectiveFunction()

[peterson-tim-j]

Currently, solve models for multiple parameter sets is primarily undertaken by parallelisation within HydroSightModel.calibrationObjectiveFunction(). This likely increases the calibration runtime because unnecessary communications is carried out and it limits the benefits from SIMD vectorization of the soil model (issue #55).

If the parrallelisation was undertaken within the model (eg model_TFN.objectiveFunction()) then vectorization of the most numerically demanding steps (ie doIRFconvolution.c and forcingTransform_soilMoisture.c) for >1 parameter sets would likely produce major improvements in calibration times. This may also require further vectorization of some functions (eg responseFunction_Pearsons.theta()).

Any thoughts on this are welcome.