powerpc64: fix handling of homogeneous float128 structs

[sharkcz]

If there is a homogeneous struct with float128 members, they should be
copied to vector register save area. The current code incorrectly copies
only the value of the first member, not increasing the pointer with each
iteration. Fix this.

[sharkcz]

Related: https://bugzilla.redhat.com/show_bug.cgi?id=2045797

Without the fix there are 2 test failures when GCC defaults to IEEE Float128 type for long double.

...
Running ../../testsuite/libffi.call/call.exp ...
Running ../../testsuite/libffi.closures/closure.exp ...
FAIL: libffi.closures/cls_align_longdouble_split.c -W -Wall -Wno-psabi -O0 execution test
FAIL: libffi.closures/cls_align_longdouble_split.c -W -Wall -Wno-psabi -O2 execution test
Running ../../testsuite/libffi.complex/complex.exp ...
Running ../../testsuite/libffi.go/go.exp ...

		=== libffi Summary ===

# of expected passes		1466
# of unexpected failures	2
# of unsupported tests		28
make[3]: *** [Makefile:457: check-DEJAGNU] Error 1

[sharkcz]

@wschmidt-ibm, could you review it, please?

[wschmidt-ibm]

Thank you, Dan! LGTM (and I see it is merged). Much obliged for the fix!