This is only a long comment.
We clearly have $i(D) = D \setminus \{ 0 \}$ for any Bézout domain $D$, while $i(k[x, y, z]) = k \setminus \{ 0 \}$ by Krull's principal ideal theorem.
As exemplified by js21 with $D = k[x, y]$ in a comment above, the answer to OP's first question is no in general. Still, I'll provide a small glint of hope.
OP's definition can be rephrased as follows.
Rewording. The set $i(D)$ consists of the elements $a \in D \setminus \{ 0 \}$ such that $D/Da$ is a principal ideal ring (PIR).
If $a, b \in i(D)$ are coprime, e.g., $Da + Db = R$, then we have have $D/Dab \simeq D/Da \times D/Db$ by the Chinese Remainder Theorem. Thus $ab \in i(D)$. In particular the set of primes $p \in D$ such that $Dp$ is a maximal ideal of $D$ is a submonoid of $D \setminus \{0\}$ contained in $i(D)$.
Using the above rewording, we can quickly check why js21's comment answers the question in the negative. If $k$ is field, we have $k[x, y]/(x) \simeq k[y]$ which is a principal ideal domain whereas $k[x, y]/(x^2) \simeq (k[x]/(x^2))[y]$ is not a PIR (the ideal generated the images of $x$ and $y$ is readily seen not to be principal). Therefore, we have $x \in i(D)$ whereas $x^2 \notin i(D)$.
Let us give another example for which multiplicative closedness of $i(D)$ fails dramatically. Consider a regular local ring $D$ (this is necessarily a domain) of Krull dimension $2$ with maximal ideal $\mathfrak{m}$. Then $i(D) = (D \setminus \mathfrak{m}) \cup (\mathfrak{m} \setminus \mathfrak{m}^2)$, but no product of two non-units in $i(D)$ belongs to $i(D)$. (The set $i(D)$ is trivially saturated).
Side note. The answers to this MSE question show why $i(D) = D \setminus \{0\}$ when $D$ is a Dedekind domain and point to interesting references.