First of all, by noticing $\cot^2(\pi - x) = \cot^2(x)$, we can write this identity as
$$\sum_{k=1,3,5} \cot^2(\frac{2\pi k}{14}) = 5$$
By writing $\cot^2(x) = \frac{1}{\sin^2(x)} - 1$, and using symmetries of $\cos$ and $\sin$ ($\cos(x)=\cos(-x)$, $\sin(\frac{\pi}{2}-x)=\cos x$), we can write this sum as follows:
$$\sum_{k=1,3,5} \frac{1}{\cos^2(\frac{\pi k}{14})} = 8$$
If we let $a_i = \cos(\frac{\pi (2i-1)}{14}), i=1,2,3$, we can write this expression as
$$(*) \frac{(\sum_{i<j} a_i a_j)^2 - 2\prod a_i \sum a_i}{(\prod a_i)^2}$$
The 7'th Chebyshev Polynomial (of the first kind) vanishes exactly on $\cos(\frac{2k-1}{14}\pi)$, $1\le k \le 7$. Those roots are actually $\pm (a_1, a_2, a_3)$ and $0$, each is a simple root.
We can compute the polynomial recursively and find that it equals
$$T_7(x) = 64x^7-112x^5+56x^3-7x=x(64x^6-112x^4+56x^2-7)$$
We'll work with $P_7(x)=\frac{T_7(x)}{64x}$, a monic polynomial with roots $\pm(a_1,a_2,a_3)$.
This shows, by using Vieta and the symmetry of roots (it requires some manipulation on symmetric polynomials):
- $(\prod a_i)^2=\frac{7}{64}$ (by considering coefficient of $x^0$)
- $(\sum_{i<j} a_i a_j)^2 - 2\prod a_i \sum a_i = \frac{56}{64}$ (by considering coefficient of $x^2$ - this one required some computation)
So the sum $(*)$ equals $\frac{56}{64} / \frac{7}{64} = 8$, which implies your identity. $\blacksquare$
EDIT: I'll describe some of the philosophy behind the answer.
The first half - I knew I wanted to you Chebyshev polynomials in some way (because its roots are related to the expression), so I did basic manipulations that helped me use the coefficients of the Chebyshev polynomial. I didn't know apriori that there are any 'good' manipulations, but I hoped and it indeed worked out.
The second half - What I really wanted is a polynomial $Q(x)$ whose roots are $a_1,a_2,a_3$. Unfortunately, I had managed only to construct the polynomial $P_7(x)$ which equals $-Q(x)Q(-x)$. Fortunately, the coefficients of $P_7$ encode enough information about the coefficients of $Q$. Explicitly, by comparing coefficients:
$$P_7[X^k] = \sum_{i+j=k} (-1)^{1+j} Q[X^i]Q[X^j]$$
I used this for $k=0,2$ and it was enough. $k=0$ gave $P_7(0)=-Q(0)^2$, i.e. we have the product of the $a_i$! (up to sign, but we don't even need it.)
$k=2$ gave $P_7[X^2] = Q[X^1]^2-2Q[X^2]Q[X^0]$, which luckily was exactly the missing ingredient in calculating the rational expression $(*)$, so that's it.
EDIT 2: I feel that I need to expand on the "theory" of Chebyshev polynomial, because using it might scare people away.
The $n$'th Chebyshev polynomial of the first kind is the unique polynomial satisfying $T_n(\cos (\theta)) = \cos(n\theta)$, for any $\theta$. Evidently, $\cos(\frac{\pi}{2n}(2k+1))$ is a root for any $k$ - just plug $\theta = \frac{\pi}{2n}(2k+1))$. As $\deg T_n = n$ (see the next paragraph), there can be no other roots.
Why is $T_n$ necessarily a polynomial? Well, for $n=0$ we have $T_0 = 1$, and for $n=1$ we have $T_1(x)=x$. For $n=2$ we already need some trigonometry: $\cos(2\theta)=2\cos^2(\theta)-1$, so $T_2(x)=2x^2-1$. We can define $T_n$ recursively by trigonometric insights:
$$\cos(\alpha)+\cos(\beta)=2\cos(\frac{\alpha+\beta}{2})\cos(\frac{\alpha-\beta}{2})$$
$$\implies \cos((n+1)\theta) + \cos((n-1)\theta) = 2\cos(n\theta)\cos(\theta)$$
$$\implies T_{n+1}(x) + T_{n-1}(x) = 2T_{n}(x)x$$
This is how I calculated $T_7$. In practice I just used the recurrence relation $T_{n+1}(x) = 2T_{n}(x)x-T_{n-1}$ and the table here.
There are some shortcuts, since the leading coefficient of $T_n$ is $2^{n-1}$ and the last coefficient is $0$ when $n$ is odd.