Breed for Longer Life Span

This is a tricky prospect, not the least of which because the female that's contributing half the genes is going to impact lifespan as well. This would have to be done with dogs who have carefully kept pedigree data, so that we could look not just at how long the male lived, but how long his parents, grandparents, siblings, aunts did (so we can be more confident there's solid reproducible genetic foundations for longevity) and so that we could reasonably evaluate the bitch's potential longevity from her relatives' statistics, since we cannot ethically be breeding females at advanced ages.

It's worth noting that most dogs who have extensively detailed pedigree records are purebreds owned by breeders and passionate breed enthusiasts who are often not the most keen on the idea of outcrossing or - in the case of females especially - donating their dogs' valuable fertile years and limited number of litters to a program that doesn't serve to further the breed they love and are devoted to.

You would also need your potential breeding pool to be full of dogs with excellent temperaments that complement each other (because you can get some seriously unstable and even dangerous results by mixing two 'excellent' but very different temperaments that clash with each other), complementary physical structures (again, some physical traits do not combine well when bred together even if they work well separately and produce physically unsound dogs), and a very low incidence of non-fatal health problems. As RonE stated, it's not worthwhile to have a dog that lives to 25 if it has serious or dangerous behavioral issues, or if it spends a large percentage of that extended lifespan with really low quality of life due to other health or structural issues. Imagine if they can physically live to 25, but start experiencing Canine Cognitive Dysfunction before they turn ten. The breeding pool would also need extremely high genetic diversity and to be large enough that inbreeding depression isn't going to become an issue and start shortening those lifespans again.

It's probably theoretically doable, especially if you do narrow your criteria to ONLY longevity, though 25 years might still be out of reach in the end - I know a few rare dogs get there, but it's so uncommon it's difficult to say if that's reproducible through selective breeding or just some kind of fluke through environment and genetics, especially since none of the dogs currently authenticated to have lived 20+ years are related as far as I can tell. There is going to be a point where extending any species' lifespan further becomes biologically impossible. It's worth mentioning that most reputable, passionate breeders DO take longevity into account when they plan their litters, and it's one of the reasons many do look at a potential stud's or dam's siblings, aunts/uncles, cousins, etc. as well as just their direct ancestors. But there's so many other things to balance when producing exceptional dogs that it cannot reasonably be the main focus of most breeding programs.

Definitely an interesting thing to think about, though!

Selecting for longevity also works differently in different breeds. Giants like the Irish Wolfhound mentioned earlier in this thread are short-lived because they're giants - their size is pushing the biological limits of the canine genome. They spend an enormous amount of resources early in life growing an incredible amount very quickly - even in cases where they're carefully fed to promote relatively 'slow and steady' growth. At their first birthday a Great Dane might easily be 100x their birth weight - compare that to a long-lived giant like the Asian elephant, who despite putting on ~1kg a day in their first year aren't even 5x their birth weight by their first birthday (on average, from the numbers I could find). For giant dogs, even as adult there's more strain on their hearts, more on their bones and joints, more on their GI system, more on their... everything. They cannot physically or biologically obtain the average lifespan of even the longer-lived breeds we currently have and still maintain the physical traits (eg size) that make them the breed they are.

It likely is possible to improve the average lifespan of giants by a small margin, partially by discouraging the breeding of the extreme end of the size range (which many reputable breeders of giant already take into account), but we'd be looking at dogs living closer to 10 years instead of 7 or 8. And even that might require outcrossing for breeds like the Dogue de Bordeaux, whose small gene pool and hereditary heart problems drag the average life span down to a measly 5-8, shorter than even other giants.

I bet you anything that if we were able to set up a superdog longevity breeding program with a huge pool of different breeds and mixes to work with, we'd wind up with a small to medium-small dog in the end, because the current evidence suggests that - on average - breeds in this size range have the longest lifespan.