I would like to do this.
I observe that I can write $$f(x,y) = (1 - x + 2y)^3 = (2(y-2) - (x-1) + 4)^3.$$ It's easy to do this via algebra directly.
However, I'm asked to do it by computing the Taylor series. Is it correct to say that I could expand the function $$f(a,b) = (2a - b + 4)^3$$ (where $a = y-2$ and $b = x-1$) around the point $(2,1)$ to the third-order term like so: $$f(a,b) = f(2,1) + f_a(2,1)a + f_b(2,1)b + \frac{1}{2}f_{aa}(2,1)a^2 + \frac{1}{2}f_{bb}(2,1)b^2 + f_{ab}(2,1)ab + \cdots$$ (where I've neglected to write the third order term for simplicity's sake)?
Or am I missing the point of the question / is my approach incorrect?
I don't understand what you're doing, you seem to be using $f$ with two different meanings.
The function is a third degree polynomial, therefore the fourth order partials are constantly zero and the Taylor series of $f$ of order $3$ (or above) coincides with $f$ (i.e. the remainder is $0$).
Thus, for all $(x,y)\in \mathbb R^2$, $f(x,y)=f((1,2)+(x-1, y-2))=(-(x-1)+2(y-2)+4)^3$ and it's easy to get the taylor expansion with powers of $(x-1)$ and $(y-2)$.
If you must use the formula for Taylor series, in the notation at the bottom of the linked section set $\mathrm x=(\mathrm a+(x-1, y-2))$ where $\mathrm a=(1,2)$ to get
$$f(x,y)=f(1,2)+\dfrac{\partial f}{\partial x}(1,2)(x-1)+\dfrac{\partial f}{\partial y}(1,2)(y-2)+\dfrac 1{2!}\left[\dfrac{\partial ^2f}{\partial x^2}(1,2)(x-1)^2+2\dfrac{\partial ^2f}{\partial y\partial x}(1,2)(x-1)(y-2)+\dfrac{\partial ^2f}{\partial y^2}(1,2)(y-2)^2\right]+R_2(x-1, y-2),$$ where $R_2$ is a function such that $\lim \limits_{(x,y)\to (0,0)}\left[\dfrac{R_2(x,y)}{\Vert (x,y)\Vert^2}\right]=0$.