A $C^2$ piecewise Hermite interpolant and a cubic spline are one and the same!
Remember what's done to derive the tridiagonal system: we require that at a joining point, the second left derivative and the second right derivative should be equal.
To that end, consider the usual form of a cubic Hermite interpolant over the interval $(x_i,x_{i+1})$:
$$y_i+y_i^{\prime}\left(x-x_i\right)+c_i\left(x-x_i\right)^2+d_i\left(x-x_i\right)^3$$
where
$$\begin{align*}c_i&=\frac{3s_i-2y_i^\prime-y_{i+1}^\prime}{x_{i+1}-x_i}\\
d_i&=\frac{y_i^\prime+y_{i+1}^\prime-2s_i}{\left(x_{i+1}-x_i\right)^2}\\
s_i&=\frac{y_{i+1}-y_i}{x_{i+1}-x_i}\end{align*}$$
and $\{y_i^\prime,y_{i+1}^\prime\}$ are the slopes (derivative values) of your interpolant at the corresponding points $(x_i,y_i)$, $(x_{i+1},y_{i+1})$.
Take the second derivative of the interpolant over $(x_{i-1},x_i)$ evaluated at $x=x_i$ and the second derivative of the interpolant over $(x_i,x_{i+1})$ evaluated at $x=x_i$ and equate them to yield (letting $h_i=x_{i+1}-x_i$):
$$c_{i-1}-c_i+3d_{i-1}h_{i-1}=0$$
Replacing $c$ and $d$ with their explicit expressions and rearranging yields:
$$h_i y_{i-1}^{\prime}+2(h_{i-1}+h_i)y_i^{\prime}+h_{i-1} y_{i+1}^{\prime}=3(h_i s_{i-1}+h_{i-1} s_i)$$
which can be shown to be equivalent to one of the equations of your tridiagonal system when $h$ and $s$ are replaced with expressions in terms of $x$ and $y$.
Of course, one could instead consider the cubic interpolant in the following form:
$$y_i+\beta_i\left(x-x_i\right)+\frac{y_i^{\prime\prime}}{2}\left(x-x_i\right)^2+\delta_i\left(x-x_i\right)^3$$
where
$$\begin{align*}\beta_i&=s_i-\frac{h_i(2y_i^{\prime\prime}+y_{i+1}^{\prime\prime})}{6}\\\delta_i&=\frac{y_{i+1}^{\prime\prime}-y_i^{\prime\prime}}{6h_i}\end{align*}$$
Doing a similar operation as was done for the Hermite interpolant to this form (except here, one equates first derivatives instead of second derivatives) yields
$$h_{i-1} y_{i-1}^{\prime\prime}+2(h_{i-1}+h_i)y_i^{\prime\prime}+h_i y_{i+1}^{\prime\prime}=6(s_i-s_{i-1})$$
which may be the form you're accustomed to.
To complete this answer, let's consider the boundary condition of the "natural" spline, $y_1^{\prime\prime}=0$ (and similarly for the other end): for the formulation where you solve the tridiagonal for the second derivatives, the replacement is straightforward.
For the Hermite case, one needs a bit of work to impose this condition for the second derivative. Taking the second derivative of the interpolant at $(x_1,x_2)$ evaluated at $x_1$ and equating that to 0 yields the condition $c_1=0$; this expands to
$$\frac{3s_1-2y_1^\prime-y_2^\prime}{x_2-x_1}=0$$
which simplifies to
$$2y_1^\prime+y_2^\prime=3s_1$$
which is the first equation in the tridiagonal system you gave. (The derivation for the other end is similar.)
The short answer is yes, if you are doing $n$'th order interpolation, you need $n+1$ points. Simpson's rule works for any function $f$. It sounds like you're doing cubic interpolation, which means you need four points. In general, the error in Simpson's rule is proportional to $f^{(n+1)}(\xi)$, where $\xi$ is some point in your interval $[a,b]$. Thus, if you're integrating a cubic polynomial via Simpson's rule, it's oppertunistic to use the third order (3/8'ths) rule because $f^{(4)}(\xi)$ for a cubic polynomial is zero for any $\xi$, hence the answer is exact.
Best Answer
If your cubic spline contains multiple segments, that means it is piecewise continuous. Each segment of the cubic spline is actually a different cubic polynomial and it is probably only C1 continuous (at best C2) at the segment joints. So, when you use Simpson's rule with n=2, all the 3 evaluations of the function value could be done against different cubic polynomials. That is the reason you do not get the same value as when using n=20. You can try to use the 3-point Simpson rule (i.e., n=2) for each segment in the spline then sum up all the result.