Fix vignette "Plugging in new calibration algorithms in airGR"

The starting points used for the multi-start approach are in the real space, while they should be in the transformed space.

The line

startGR4J <- expand.grid(data.frame(CalibOptions$StartParamDistrib))

should be replaced by

StartParamDistrib <- matrix(c(+5.13, -1.60, +3.03, -9.05,
                         +5.51, -0.61, +3.74, -8.51,
                         +6.07, -0.02, +4.42, -8.06), ncol = 4, byrow = TRUE)
startGR4J <- expand.grid(data.frame(StartParamDistrib))

or perhaps, by a more diverse list of starting points sampled in the transformed space if we want to find local optima.