Let $a \in \Bbb{F}_p$. $a\not\in (\Bbb{F}_p^{\times})^2$ and $b\not\in (\Bbb{F}_p^{\times})^2$ implies $ab\in (\Bbb{F}_p^{\times})^2$ because $\left (\frac{a}{p}\right)\left (\frac{b}{p}\right)=\left (\frac{ab}{p}\right)$ holds.
Let $a \in \Bbb{F}_q$. But does $a\not\in (\Bbb{F}_q^{\times})^2$ and $b\not\in (\Bbb{F}_q^{\times})^2$ imply $ab\in (\Bbb{F}_q ^{\times})^2$, where $q=p^r$($r \ge 2$ is positive integer) ?
I have never heard generalization of Legendre symbol into general finite field, so this may have some counterexamples.
Yes. As the homomorphism map $x\mapsto x^2$ from $\mathbb F_q$ to itself has kernel $\{\pm 1\}$, we know the image has precisely $|\mathbb F_q^{\times}|/2$ elements (assuming $2\nmid q$, otherwise the problem is trivial). In other words, $\mathbb F_q^{\times}/(\mathbb F_q^{\times})^2\simeq \mathbb Z/2\mathbb Z$ as abelian groups. Hence if $a,b\not\in(\mathbb F_q^{\times})^2$, in other words, their images in the quotient group both correspond to $1$ in $\mathbb Z/2\mathbb Z$, then $ab$ must correspond to $0$, that is, $ab$ is in the kernel $(\mathbb F_q^{\times})^2$.
Put in another way, $\mathbb F_q\rightarrow\mathbb F_q^{\times}/(\mathbb F_q^{\times})^2\simeq (\{\pm 1\}, \text{multiplication})$ is precisely the generalization of the Legendre symbol.