Since the matrices that do not commute is not a subspace, we need to take an approach different from the one you outline in your last paragraph. A better approach is the one given by @Arthur in his comment to the question. We want to find all matrices $B$ where $AB=BA$. We let
$$B = \begin{bmatrix}
a & b \\
c & d \\
\end{bmatrix}$$
We then want
$$\begin{bmatrix}1 & 2 \\2 & 1 \\\end{bmatrix} \begin{bmatrix}a & b \\c & d \\\end{bmatrix}
=\begin{bmatrix}a & b \\c & d \\\end{bmatrix} \begin{bmatrix}1 & 2 \\2 & 1 \\\end{bmatrix}$$
$$\begin{bmatrix}a+2c & b+2d \\2a+c & 2b+d \\\end{bmatrix}
=\begin{bmatrix}a+2b & 2a+b \\c+2d & 2c+d \\\end{bmatrix}$$
Then the upper left corners give us
$$a+2c=a+2b$$
$$b=c$$
The upper right corners give us
$$b+2d=2a+b$$
$$a=d$$
This means that
$$B = \begin{bmatrix}a & b \\b & a \\\end{bmatrix}$$
Checking, we get
$$AB=\begin{bmatrix}a+2b & 2a+b \\2a+b & a+2b \\\end{bmatrix}=BA$$
so that answer is correct. We see that
$$B = a\begin{bmatrix}1 & 0 \\0 & 1 \\\end{bmatrix}+b\begin{bmatrix}0 & 1 \\1 & 0 \\\end{bmatrix}$$
Therefore we conclude that the matrices that commute with $A$ are a two-dimensional subspace of the vector space of $2$-by-$2$ matrices, with the basis
$$\left\{\begin{bmatrix}1 & 0 \\0 & 1 \\\end{bmatrix}, \begin{bmatrix}0 & 1 \\1 & 0 \\\end{bmatrix}\right\}$$