The step where you go from $\lim_{n\to\infty}\left(1+\frac{1}{n}\right)^n(n+1)- \left(1+\frac{1}{n-1}\right)^n(n-1)$ to $\lim_{n\to\infty} e(n+1)-e(n-1)$ is invalid. You can't evaluate just some pieces of the limit in terms of $n$. Formally, you're starting with an indeterminate form of type $\infty - \infty$.
As a simpler example as to why your reasoning doesn't work, consider the indeterminate form $(n+1)-n$, which is identically equal to $1$. But it can be rewritten to be of the form
$$
n\left(\frac{n+1}{n}\right) - (n+1)\left(\frac{n}{n+1}\right)
$$
Now both $\left(\frac{n+1}{n}\right)$ and $\left(\frac{n}{n+1}\right)$ converge to $1$, but the limit of $n(1) - (n+1)(1)$ is now $-1$ instead of $1$.
As to how to derive the identity, the Mathworld link suggests that a reasonable approach would be to start with the expression
$$
\left(1 + \frac{1}{x}\right)^x
= e\left[1 - \frac{1}{2x} + \frac{11}{24x^2} - \frac{7}{16x^3} + \frac{2447}{5760x^4} - \dots\right]\, ,
$$
which is derived by using a composition of the identity
$$
x\ln\left(1 + \frac{1}{x}\right)
= 1 - \frac{1}{2x} + \frac{1}{3x^2} - \frac{1}{4x^3} + \dots
$$
with the power series expansion of $e^x$.