[RFC] Constructor patch

[Tobias Burnus]

Damian Rouson wrote:
> Thanks for copying me on this.  I'm excited to see the progress. A
> test case is attached to this e-mail.

Works. The commented line,
!  a = mycomplex(argument=5.6, modulus=1.0)
only worked after I also removed the "private" from the 
mycomplex_module, which was to be expected.

The test is nice as it works -- and as I had actually forgotten that one 
can have also a program where the generic procedure and the structure 
constructor have an ambiguous interface, which is only distinguished by 
the keywords=.

Tobias

PS: I have now fixed the IMPORT issue, which brings down the failures 
from 51 to 29. I assume some of  the remaining failures will be more 
difficult, but let's see...