chore: adopt better keycmp() implementation

[Integral-Tech]

According to the source code of strcmp() in glibc, the return value is the difference of first different character in two strings.

For keycmp(), we can adopt an implementation akin to strcmp() in glibc.

[eafer]

Have you tested this code? Because in C, substraction of unsigned types will wrap around and remain positive. The glibc version works because unsigned char gets promoted to int beforehand. Yes, this is all very messy.

[Integral-Tech]

@eafer

      unsigned char c1 = *p1++;
      unsigned char c2 = *p2++;
      int diff = c1 - c2;

I don't think unsigned char gets promoted to int beforehand in glibc strcmp().

[eafer]

The promotion is implicit.

[Integral-Tech]

@eafer IMO this code is akin to strcmp() in glibc, as the return value is int.

static int omap_keycmp(struct apfs_omap_key *k1, struct apfs_omap_key *k2)
{
	if (le64_to_cpu(k1->ok_oid) != le64_to_cpu(k2->ok_oid))
		return le64_to_cpu(k1->ok_oid) - le64_to_cpu(k2->ok_oid);
	if (le64_to_cpu(k1->ok_xid) != le64_to_cpu(k2->ok_xid))
		return le64_to_cpu(k1->ok_xid) - le64_to_cpu(k2->ok_xid);
	return 0;
}

[eafer]

The cast only happens after all the unsigned math is done, so the result could wrap around so far that it becomes a positive int.