A Ruffini radical in this context means that it isn't allowed to be of a form so that it "brings in" any radicals (n-th roots) from anywhere else than the field generated by the roots (and the polynomial coefficients, which is a subfield). The non-Ruffini radical fields that Stewart implicitly talks about don't have that limitation but can be pulled from anywhere. They don't have the limitation $K_j \subseteq \mathbb{C}(t_1,...,t_n)$.
Stewart says on page 118 in the 4th edition (about Ruffini radicals): "The aim of this definition is to exclude possibilities like $\sqrt{-121}$ in Cardano's solution of the quartic equation $t^4-15 t-4=0$, which does not lie in the field generated by the roots, but is used to express them by radicals."
A section or two forward in the book, after the mentioning of Ruffini radicals, Stewart has included a theorem and proof of Abel (difficult) that shows that as a matter of fact bringing in roots from "anywhere else" doesn't bring any more power to the game. So Ruffini radicals turns out to be all you need. This blog talks about this: https://jmanton.wordpress.com/2016/08/11/a-snippet-of-galois-theory/
Doing Abel's proof using Galois theory (which didn't exist when Abel did his proof) is much easier and is covered by, for example, Wikipedia: https://en.wikipedia.org/wiki/Abel%E2%80%93Ruffini_theorem
Stewart shows how to do it "the hard way" without Galois theory, the way Abel did it. This is what makes it necessary to introduce "Ruffini radicals".
(Note that some editions of Stewart's book seem to have a typo. In the definition of Ruffini solubility it should say $K_{j+1}=K_j(α_j)$)