We have
$$\left|x^2-x\over x-8\right|=|x|\left|x-1\over x-8\right|\le|x|{|x|+1\over||x|-8|}$$
Now the main thing that causes a problem is the smallness of the denominator if $|x|\approx8$. But since we're interested in the limit as $x\to0$, it's easy to stay away from $8$ by requiring, say $|x|\le7$, in which case we have
$${|x|+1\over||x|-8|}\le{7+1\over|7-8|}=8$$
So taking $\delta=\min(7,\epsilon/8)$ we have
$$0\lt|x|\lt\delta\implies\left|x^2-x\over x-8\right|\le8|x|\lt8\delta\le\epsilon$$
as desired.
Remark: The key to limit write-ups of this type is to keep in mind that even though we're thinking of small epsilons, we don't want to be embarrassed by claiming an implication that isn't true in case someone decides to use a large $\epsilon$. E.g., if we just let $\delta=\epsilon/8$, we run the risk of the false implication $0\lt|x|\lt8\implies|(x^2-x)/(x-8)|\lt64$, which is falsified, for example, at $x=7.9$ (that is, $0\lt|7.9|\lt8$ but $|(7.9^2-7.9)/(7.9-8)|=545.1\gt64$). The role of the "min" is to avoid such false claims. As the other answers here show, there is no single prefered way to choose your delta; it's somewhat a matter of taste. My preference is to try to isolate what causes things to be uncontrollably large, usually a potential zero in a denominator, impose a restriction brings that term under control, and then look at what happens to everything else.