You seem to have gone off-track. You are quite correct that since $G$ is $G_\delta,$ then there exists a family of open sets $\{\mathscr{O}_k: k \in \mathbb{N}\},$ and I assume that you are wanting to say that the intersection of all of these open sets is exactly $G.$ Unfortunately, that's not what you've said. Rather, you should say that $G=\bigcap_{k\in\Bbb N}\mathscr O_k.$ Then we easily have $$G+y=\left(\bigcap_{k\in\Bbb N}\mathscr O_k\right)+y,$$ but we should justify that this is equal to $\bigcap_{k\in\Bbb N}(\mathscr O_k+y)$.
Fortunately, it isn't too difficult to prove that if $\mathcal A$ is a set of subsets of $\Bbb R,$ then for any $y\in\Bbb R,$ we have $$\left(\bigcap_{A\in\mathcal A}A\right)+y=\bigcap_{A\in\mathcal A}(A+y).$$
Now, your claim that $G+y$ is open is unfounded. After all, $\{0\}$ is a $G_\delta$ set, and no translate of $G$ is open! Rather, you need to justify that $\mathscr{O}_k+y$ is open for all $k\in\Bbb N.$ It is enough to prove that the translate of an open interval is open. (Why?) Thus, as a countable intersection of open sets, $G+y$ is a $G_\delta$ set, as desired.
All that is left is to prove that if $A\subseteq\Bbb R$ and $y\in\Bbb R,$ then $(A+y)^c=A^c+y,$ whence we can apply DeMorgan's Laws together with the previous results to get the rest.
Edit: It looks better! Again, I must note that $G+y$ is not (necessarily) an open set, but a $G_\delta$ set.
Regarding your proof that $I+y$ is open: you should show (unless you're relying on previous results) that for any $x\in I+y,$ there exist $c,d\in I+y$ such that $x\in(c,d).$ Use openness of $I$ to find $c',d'\in I$ such that $x-y\in(c',d'),$ and proceed.
Regarding your proof that $A^c+y=(A+y)^c,$ we should proceed in a related fashion. Since $x\in A^c+y$ iff $x-y\in A^c$ iff $x-y\notin A$ iff $x\notin A+y$ iff $x\in(A+y)^c$ (which is fairly straightforward to justify), then we're done. Also, "$a+y^c$" doesn't really make sense--what would be the complement of a number?
It looks good, otherwise (from what these tired eyes can see).
Best Answer
A correct and rigorous proof of this statement you can find it in Royden's Real Analysis in chapter 2 theorem 11. The argument you are asking is just almost the proper definition of the Lebesgue outer measure, remember that outer measure if defined as the inf{of possible outer coverings by open sets}, then all follows from the definition of inf, the monotonicity of Lebesgue outer measure and the excision property of measurable sets.