misleading warning on sriv fit when using inverse fit

[josephguillaume]

Failed SRIV fits may not be important when calling inverse fit, if they occur early in the iterative fitting process. The fallback method modifies U such that later iterations of SRIV succeed. The warnings produced by hydromad at the moment are misleading, and can only by diagnosed by setting: hydromad.options(trace=TRUE) and options(warn=1)