Griffiths puts in the approximation signs $\approx$ because the argument is generally handwavy, but the inner relation is actually exact:
$$h(\xi) \approx C \sum \frac 1 {(j/2)!} \xi^j = C \sum\frac 1 {j!} \xi^{2j} = C e^{\xi^2}.$$
The reason for this is that the sum over $j$ is in steps of two, because the symmetry of the equation allows you to have separate even and odd series. The inner step is just a relabelling of the series, and the final step is a simple calculation.
The real mettle of the approximation is saying that the recursion of the series,
$$
a_{j+2}=\frac{2j+1-K}{(j+1)(j+2)}a_j
$$
can be reduced to the approximate form $a_{j+2}\approx \frac2j a_j$ at large $j$. If you're more precise then you use that intuition to show that, if you define $a_j$ as above with (say) $a_0=1$ and $a_1=0$, and you define as a standard $b_{j+2}=\frac2j b_j$ with $b_0=1$ and $b_1=0$, then there exist an index $J$ and a constant $c$ such that for all $j\geq J$ you have $a_j>c b_j$, i.e. the series you're interested in is bounded below by your standard, which produces a divergent function. Once you show that, then it follows that
$$
h(\xi)>\sum_{j=0}^{J-1}(a_j-b_j)\xi^j+c\sum_{j=0}^{J-1}b_j\xi^j
=\sum_{j=0}^{J-1}(a_j-b_j)\xi^j+ce^{\xi^2}
$$
which diverges at large $\xi$.