New Wigner code for inclusion

["Jonathan G. Underwood" <j dot underwood at open dot ac dot uk>]

Dear All,

Following on from my previous email regarding code for replacing that in 
coupling.c [see previous email to this list for the reasons] as well as 
adding the computation of reduced wigner rotation matrix code to the 
library, I've now implemented error approximation according to the rules 
generally used for the other specfunc functions. Gerard, if you have a 
spare moment, this could definately benefit from having your expert eye 
cast over it to check the error analysis is roughly in line with other 
functions. I've also added documentation and testing code. See the files at

http://www.physics.open.ac.uk/~ju83

Specifically:
wigner.c the new code, and gsl_sf_wigner.h the accompanying header file.

test_sf.c - taken from gsl-1.5, with added code for testing wigner.c, 
and comments for the test of coupling.c where incorrect values are 
returned for undefined symbols (where 0 and not NaN is returned)

specfunc-wigner.texi documentation
specfunc.texi taken from gsl-1.5 and added the node for the new 
documentation.

This code is now hopefully ready for inclusion in gsl. It can co-exist 
alongside coupling.c, but I recommend we remove that and, if necessary, 
add macros for the _coupling_ functions to point at the wigner functions 
for backward compatibility. If you would like me to cook up some 
patches, I'm happy to do so.

Best wishes,
Jonathan