Work on the bug in the Incomplete Gamma function

As I have already reported on this mailing list I am working on the bug
in the Incomplete Gamma function. I have finished a more carefully
implementation of the regions in the complex plane where we have to use
the series or the continued fraction expansion to calculate the
numerical values.

Now I am doing more tests to be sure that it will work. Especially I am
testing the function on the boundaries were we change from the series
expansion to the continued fraction expansion. At these boundaries both
algorithm should give the same results within an expected numerical
precision.

Dieter Kaiser