regression in subtypep wrt complex type specializers
Related to fixes in #468 (closed) (most notably not a regression in complex-float feature branch). Regression revealed by ansi-tests
> (subtypep '(complex fixnum) '(complex integer)) NIL T CL-USER> (subtypep 'fixnum 'integer) T T > (subtypep '(complex integer) '(complex fixnum)) T T CL-USER> (subtypep 'integer 'fixnum) NIL T
Subtypep seems to give reversed results for complex types. Problem does not occur in the last release.