I've been making bits of this argument for years and maybe I should put it together. Here goes:
1. It is rational to trade off longevity for other kinds of benefits. Actually, it is not only rational, but commonly done. As Gary Becker says, most deaths are suicides: if one's overriding objective in life were to live as long as possible, one would live quite differently from the way people actually do.
2. Different people will presumably want to make their longevity-happiness tradeoff in different ways; there are no objectively right or wrong ways to do this; in the absence of externalities and other compelling reasons, the state should let them make such tradeoffs for themselves.
3. In a society with universal healthcare (e.g., Britain) it is sometimes claimed that fat people are a burden on the healthcare system. This isn't actually true. Fat people cost more per year of life, but less overall. (e.g. This study shows that the lifetime medical expenses of the obese/smokers are less than those of "fit" people.) Remember, the life expectancy even for obese people is somewhat higher than the retirement age. Therefore, healthy people generally spend their additional years of life as parasites rather than as generators of wealth. [In a society like the US the situation is unjust to the fat. Obese people, smokers, and others who are less likely to live past 65 are subsidizing the medical expenses of those who live past 65. There seems to be no compelling economic reason to provide this kind of incentive for living past 65.]
4. Life expectancy is not a sound measure of welfare any more than GDP per head. As Krugman argues here, the French put up with earning less because they get to take longer vacations. Similarly, obese alcoholics might be fine with heart attacks because they get to drink bacon martinis while they are alive. These are both, in principle, rational tradeoffs, though in both cases you can imagine conditions (acquiring expensive new obligations, being on your deathbed) where people regret their rationally made choices.
5. Exceptions should, I suppose, be made for the case of serious physical addiction, in which the addict seriously wants to change his lifestyle but can't. For long-time smokers I imagine this is a substantial fraction of the relevant demographic, but for drinkers, stoners, compulsive eaters, and non-exercisers it's a relatively small segment. e.g. there's a large grey area between drinking enough to increase your chances of liver cancer and drinking enough to be unable to go without booze; this is presumably where most regular drinkers fall.
6. Unlike smoking or public drunkenness, obesity does not obviously hurt other people. (Except on airplanes. I don't really have a problem with charging people extra if they don't fit in their seats.)
7. Arguably being obese increases your chances of being unemployed and therefore on public assistance. (And maybe it increases your chances of being poor and hence a beneficiary of progressive tax credits and/or welfare policies.) I'm pretty sure, however, that the causation goes the other way, and that whatever residual effect there might be is largely explained by anti-fat discrimination. If one could establish that being fat or a smoker made you sufficiently unproductive per hour that you were on the whole a ward of the state in times of full employment I would probably change my mind on this entire issue.
So to sum up etc. (1) I don't see that society has a compelling reason to want people to make the health/pleasure tradeoff in a way that's maximally skewed toward health. (2) One could argue that by feeding kids healthy food when they're young you avoid the tradeoff altogether because they learn to enjoy the mashed yeast. I am deeply skeptical of this: in general, food can be arbitrarily healthy and arbitrarily good for you only to the extent that the ingredients are arbitrarily upscale. This is unworkable for a large number of people. (The usual ways of dealing with lousy ingredients, e.g. spicing the hell out of everything a la Indian food, come with health costs.)
I should note that the basic reason I care about this issue is that I think Megan McArdle's right that there's something very Victorian and moralistic about the (blue-state) public's attitude to fat. I find this sort of hardworking/lazy dichotomy extremely irritating esp. since the "upright" classes tend to be smug, rich, and liberal.