Solution to Part (1)
If $P(a)=0$ and $P'(a)=0$, then $\left(a^2+a-1\right)^n=a$ and $n\,(2a+1)\,\left(a^2+a-1\right)^{n-1}=1$. Consequently, $$n\,(2a+1)\,a=n\,(2a+1)\,\left(a^2+a-1\right)^n=1\cdot \left(a^2+a-1\right)\,.$$ That is,
$$(2n-1)\,a^2+(n-1)\,a+1=0\,.$$
Since $\left(a^2+a-1\right)^n=a$, we see that $a$ is an algebraic integer. Therefore, the monic minimal polynomial of $a$ over $\mathbb{Q}$ is in $\mathbb{Z}[x]$. Hence, there is a monic polynomial $f(x)\in\mathbb{Z}[x]$ dividing the quadratic polynomial $g(x):=(2n-1)\,x^2+(n-1)\,x+1$. Consequently, either that the leading coefficient $2n-1$ of $g(x)$ divides its constant term $1$ (leading to $n=1$), or that $f(x)$ is linear (so $a\in\mathbb{Z}$, yielding $a\in\{-1,+1\}$). In the latter case, we have $$
1=n\,(2a+1)\,\left(a^2+a-1\right)^{n-1}=\left\{
\begin{array}{ll}
3n\,,&\text{if }a=+1\,,\\
(-1)^n\,n\,,&\text{if }a=-1\,,
\end{array}\right.$$
which is absurd. Thus, $n=1$ is the only possibility, but then $P(X)=X^2-1$ has no double roots.