This theorem is given in the book.
Statement :
Let $$ax^2 + by^2 + 2hxy + 2fy + 2gx + c = 0$$ be a equation of a curve and let $my + nx = l, l \ne 0$ be a line intersecting the given curve at two points, then find the equation of pair of lines joining origin and the points of intersection of given curve and given line. (mind-bending statement).
Proof :
$$my + nx = l \implies {my + nx \over l} = 1$$
Therefore it easily follow from here that the desired equation is $$ax^2 + by^2 + 2hxy + 2(fy + gx)\left({my + nx \over l}\right) + c\left({my + nx \over l}\right)^2 = 0 \tag{P}$$
- Please help me understand from where (P) come from ? Also why we are discriminating $ax^2, by^2$ and $2hxy$ by not multipling them by $\displaystyle {my + nx \over l}$ ?
The author is simply expanding $1$ in the formula. You could use this as an intermediate step:
$$ ax^2+by^2+2hxy+2(fy+gx)\cdot 1 + c\cdot 1^2=0 $$
The more interesting question is that of why they are doing this. The key point is that this substitution makes all the terms quadratic. If you expand these, you get something of the form
$$\alpha x^2+\beta y^2+\gamma xy=0$$
with $\alpha,\beta,\gamma$ rational functions in $a,b,c,d,e,f,g,l,m,n$. Now the origin $x=y=0$ is definitely a solution of this equation. And the points of intersection will be solutions, too, since they satisfy both the constituent equations.
You can also see that if $(x,y)$ is a solution, then $(\lambda x,\lambda y)$ is a solution, too. So if you have a solution distinct from the origin, then the whole line joining that solution to the origin will consist of solutions, too.
That would raise the combined degree of $x$ and $y$ to more than two in parts of the formula. The result would be an algebraic curve which does contain the origin and the two points of intersection. But it would be more complicated, not simply a pair of lines. The argument about $\lambda$ I made above no longer holds in this case.