rs6000: sysv: Don't pass SFmode in varargs in FPRs

rs6000: sysv: Don't pass SFmode in varargs in FPRs

This makes the float32-basic.c testcase work on sysv (32-bit Linux).

"float" is promoted to "double" for varargs.  The ABI also only defines
the use of double precision in varargs.  But _Float32 is not promoted.
Since there is no way of passing single-precision float in FPRs we
should pass SFmode in GPRs (or memory) instead.  This is similar to
the 64-bit ABI.

From-SVN: r258454