Two solutions. The first is based on resultants. As I wasn't 100% confident in my handling of the leading coefficients, I calculated the discriminant also using the definition. I am keeping the first (worse) solution for "educational purposes".
A promising way is to use the description of the discriminant as a resultant of $f(x)$ and $f'(x)$. The resultant is (when non-zero) also a generator of the ideal $I=(f(x),f'(x))\cap\mathbb{Z}$ and can be calculated with steps like the Euclidean algorithm, where we need to take care to keep everything as integers (necessitating appropriate scaling).
Here we first calculate
$$
r_1(x)=nf(x)-xf'(x)= -b(n-1)x +cn
$$
and are "lucky" to get a linear remainder. Proceeding we see that
$$
g(x)=(cn)^{n-1}-[b(n-1)x]^{n-1}
$$
is a multiple of $r_1(x)$ by a polynomial in $\mathbb{Z}[x]$. We can then eliminate $x$ by calculating the combination
$$
r_2(x)=(b(n-1))^{n-1}f'(x)+ng(x)=n^nc^{n-1}-(n-1)^{n-1}b^n.
$$
Here's a more convincing calculation. Let us write
$$
\begin{aligned}
f(x)&=(x-\alpha_1)(x-\alpha_2)\cdots (x-\alpha_n),\\
f'(x)&=n(x-\beta_1)(x-\beta_2)\cdots (x-\beta_{n-1}).
\end{aligned}
$$
Here the numbers $\beta_j, j=1,2,\ldots,n-1,$ are the zeros of $f'(x)$, i.e. we have $\beta_j=\zeta^j K$, where $K=\root{n-1}\of {b/n}$ and $\zeta=e^{2\pi i/(n-1)}$ is an appropriate root of unity.
Now
$$
\begin{aligned}
\operatorname{disc}(f)&=\prod_i f'(\alpha_i)=n^n\prod_{i,j}(\alpha_i-\beta_j)\\
&=n^n(-1)^{n(n-1)}\prod_{i,j}(\beta_j-\alpha_i)\\
&=n^n\prod_jf(\beta_j).
\end{aligned}
$$
We can calculate
$$
f(\beta_j)=\beta_j(\beta_j^{n-1}-b)+c=c-(b-K^{n-1})\beta_j=c-\frac{b(n-1)K}n\zeta^j.
$$
so
$$
\operatorname{disc}(f)=n^n\prod_j(c-\frac{b(n-1)K}n\zeta^j).
$$
We can use the factorization
$$
u^{n-1}-v^{n-1}=\prod_j(u-\zeta^jv)
$$
with $u=c$ and $v=b(n-1)K/n$ to evaluate this. The result is
$$
\operatorname{disc}(f)=n^n\left(c^{n-1}-\left(\frac{b(n-1)K}n\right)^{n-1}\right)
=n^n\left(c^{n-1}-\frac{(n-1)^{n-1}b^n}{n^n}\right),
$$
which equals the claimed formula.