You are right in that

    31 c0             xor %eax, %eax

is shorter than

    b8 00 00 00 00    mov $0, %eax

but that's not the main reason it's often used in shellcode. The xor reg, reg is preferred because it means the resulting shellcode will not contain a null byte. This is desirable since the null-byte is often used as a string terminator, in environment variables for example.