javaop.h type collision on AIX

[Per Bothner]

> 	I think that there are a couple of problems: signed-ness
> assumptions and not protecting all types of previous definitions.

It seems like there is a third problem:  Why are we seeing the
conflicting definition anyway?  Is AIX polluting the application
namespace, or is there something else going on?

The signed-ness problem is complicated because "signed char"
is not valid in K&R C.  We probably don't care (the Java
compiler should only be compiled using gcc), but for robustness
the following may make sense:

- Remove the definition of int8.
- Replace the definition of jbyte by:

#ifdef __STDC__ 
typedef signed char             jbyte; 
#else 
typedef char                    jbyte; 
#endif 

	--Per Bothner
Cygnus Solutions     bothner@cygnus.com     http://www.cygnus.com/~bothner