configure: only use -ffloat-store on i386

this option is expensive and only used on old gcc's that lack
-fexcess-precision=standed, but it's not needed on non-i386 archs
where floating point does not have excess precision anyway.

if musl ever supports m68k, i think it will need to be special-cased
too. i'm not aware of any other archs with excess precision.

configure

@@ -209,7 +209,7 @@ tryflag CFLAGS_C99FSE -nostdinc
 tryflag CFLAGS_C99FSE -ffreestanding \
 || tryflag CFLAGS_C99FSE -fno-builtin
 tryflag CFLAGS_C99FSE -fexcess-precision=standard \
-|| tryflag CFLAGS_C99FSE -ffloat-store
+|| { test "$ARCH" = i386 && tryflag CFLAGS_C99FSE -ffloat-store ; }
 tryflag CFLAGS_C99FSE -frounding-math
 
 #