Bug in gsl_sf_gamma_inc_P_impl

[Gerard Jungman]

Teemu Ikonen wrote:
> 
> I get GSL_EMAXITER when trying to calculate
> gsl_sf_gamma_inc_P_impl(100.0, 99.0+-eps, r)
> with gsl-0.6.
> 
> Other x values work ok, only values very close to 99.0 cause the routine
> to fail.

This turned out to be a more general problem for the case a=x+1,
but only on some paths. I fixed the continued-fraction
evaluation on the failure path, so it should be ok now.

I just checked in the fixes.

Thanks.

-- 
G. Jungman