Useless organs have no "selective pressure". So they should simply be left alone. Thus they should be accumulated in the process of evolution.
However, organs cost resources to build and maintain. Reproduction also costs resources: making sex cells, finding mates, fighting rivals and whatnot. An animal can only get so much food, which doesn't exist in infinite quantities.
So, if you have a big expensive organ that you don't use for anything, your reproductive success is going to suffer compared to those who have a smaller version or none of the useless organ at all. In these cases, there IS a selective pressure on the useless organs - in the direction of making them shrink.
The ones that can stay are the ones that confer no particular disadvantage. The ones that
will definitely stay are the ones that are so heavily integrated into the developmental program that you can't remove them without upsetting something very important.
Is nipple a definitional organ for mammals?
No.
Is it possible to milk without a nipple?
Yes, monotremes do it. The nipple doesn't produce milk, it just makes it easier for the baby to suck it.
Why should human have only two nipples rather than several?
Because they rarely have more than two offspring at a time. They rarely have even two.
I think some people are born with multiple nipples, though. *searches* Yes, it seems
it's quite common, although the additional nipples are usually not well developed. Again, a remnant of a past when our ancestors were more conventional mammals with larger litters.
Or, why not just has one?
I don't know if there's a specific advantage to having paired nipples (other than being helpful for mothers of twins), but having just one nipple would require symmetry breaking, and my guess is that that's not the easiest thing to evolve. (If anyone has more than a guess, though, I wouldn't mind some education)
If so, why did it appear at all? You (?) gave me an impression that the first mammal did NOT have mammary gland.
It
had mammary glands, it probably didn't have nipples. The two are not the same thing.
As I said, the most obvious reason I can think of is that nipples make suckling more efficient. A baby mammal sucking a nipple can probably (I haven't seen actual measurements) get milk faster than one licking milk off mum's hair, and I'm pretty sure that less milk goes to waste if there's a nipple. I think it's a reasonable assumption that that translates into better growth for the young at a lower price for the mother.
Also, why do mammals need tailored food like milk? It increases the dependence of baby and is unfavorable to evolutional process.
Not necessarily.
First, parental care in general has a number of advantages:
(1) It increases the chance that any particular offspring will survive. It allows for fewer offspring overall (imagine having to care for the millions of eggs some fish lay!), but most of the offspring of non-caring species die very young, so it balances out.
(2) It lets the young grow faster: while the offspring of non-caring species must spend a lot of time and energy finding food, the young of many caring species don't have to do anything but sit in one place and eat what their parents bring (in this respect, many insects count as "caring" species, since they lay eggs in the middle of a huge, often protected, food supply). And grow. Faster growth is good for more than one reason: first, it gets you out of the "everyone eats you" size range more quickly, and second, it means you can mature and reproduce earlier.
Milk in particular is good because again, a food source tailored to your needs allows better growth than a food source that doesn't have an ideal composition. Plus, milk is a great way of immunising a newborn against many diseases while its own, previously sheltered, immune system is still learning to recognise pathogens.
So no, milk isn't "unfavorable to evolutional process". It's just one of several alternative strategies ensuring that at least a few of your genes get into the new generation, and seeing as there are still a few thousand mammalian species all over the globe a couple hundred million years (or more) after the invention of milk, it clearly can't be that bad.