Because there's no advantage in having the same calling convention. Go code and C code cannot call each other directly even when the calling convention would be the same because Go uses split stacks.
OTOH, it makes sense in gccgo, as gcc supports C split stacks for some architectures. And, IIRC, there the calling convention is because of that compatible. (More details here.)
Disclaimer: I didn't ever actually used gccgo.