When decoding a base-{16,32,64} value, clear the target buffer first
This is a good idea in case the caller stupidly doesn't check the
return value from baseX_decode(), and as a workaround for the
current inconsistent API of base16_decode.
Prevents any fallout from bug 14013.