Why can't the quadratic formula be simplified to $x=\frac{-b\pm\sqrt{b^2-4ac}}{2a}=\frac{-b\pm(b-2)\sqrt{ac}}{2a}$?
Because $\sqrt{b^2 - 4ac}\neq 2b\sqrt{ac}$. Say for example that $c = 0$ and $b\neq 0$. Then you have \begin{align*} \sqrt{b^2 - 4ac} &= \sqrt{b^2}\\ &= \left|b\right|\\ &\neq 0, \end{align*} so you can see that this simplification cannot be correct.
It seems that in your proposed simplification, you have completely disregarded the subtraction occurring in the radical. Moreover, $\sqrt{b^2} = \left|b\right|$, not just $b$. To see this with an example, take $b = -1$. Then $\sqrt{(-1)^2} = \sqrt{1} = 1 = \left|-1\right|\neq -1$.
EDIT:
Again, the simplification is incorrect. While it is true in general that $\sqrt{a^2b} = \left|a\right|\sqrt{b}$, this is not the situation you are in here: $$ b^2 - 4ac\neq (b - 2)^2 ac = (b^2 - 4b + 4)ac = b^2 ac - 4abc + 4ac. $$ You seem to have made a few mistakes here (if I'm to take a stab at the reasoning behind the simplification): first you've incorrectly simplified $b^2 - 4ac$ as $(b^2 - 4)ac$ (which is not true, because the first term in the former has no $ac$), and then you've simplified $b^2 - 4$ as $(b - 2)^2$, which is also not true (take $b = 0$ to see why). In general, $(x + y)^n\neq x^n + y^n$: this is a common mistake algebra learners make! Remember that when expanding $(x + y)^2$, we need to use the distributive property, and not simply regard squaring as linear: \begin{align*} (x + y)^2 &= (x + y)(x + y)\\ &= x^2 + yx + xy + y^2\\ &= x^2 + 2xy + y^2. \end{align*}
The "simplification" is incorrect: $\sqrt{b^2-4ac}\not=2b\sqrt{ac}$.
For example, take $b=1, a=c=0$. Then the former expression is $1$ but the latter is $0$.
What is true is that $2\vert b\vert\sqrt{ac}=\sqrt{b^2\cdot4ac}$, but note the replacement of "$-$" with "$\cdot$", there: that's a major change! (Also note the absolute value, which is important but less fundamental in this case.)
EDIT: the question has now been changed to reflect a new simplification - namely, $$\sqrt{b^2-4ac}=(2-b)\sqrt{ac}.$$ However, this one is also false: again, set $a=c=0$, $b=1$ to see the difference.
Note that this example really shows that the discriminant can't (in general) be written in the form $[stuff]\sqrt{ac}$. Note that this applies to the new edit, which replaced "$b-2$" with "$2-b$" (as well as the very first version) - the "simplification" is still wrong, for the same reason. In a certain sense, any expression of the form you are looking at gives $a$ and $c$ "too much power" over the value; no choice of $a$ and $c$ can guarantee that the discriminant is zero regardless of what $b$ is.
This time it's not clear to me what the algebra error is; can you explain why you thought this simplification worked? EDIT: Stahl's answer takes a stab at guessing what happened; if that's not an accurate interpretation, please explain how you came by this "simplification."
FURTHER EDIT: You've added your reasoning; you make two fundamental mistakes. The gist of your argument is $$\sqrt{b^2-4ac}=\sqrt{b^2}-\sqrt{4ac}=(b-2)\sqrt{ac}.$$ Both of these equalities are false.
In the first case, we do not have $\sqrt{X+Y}=\sqrt{X}+\sqrt{Y}$, any more than we have $(X+Y)^2=X^2+Y^2$. For an explicit counterexample, take $X=Y=2$, where $\sqrt{X+Y}=\sqrt{4}=2$ but $\sqrt{X}+\sqrt{Y}=2\sqrt{2}>2$.
For the second one, it is true that $\sqrt{b^2}-\sqrt{4ac}=b-2\sqrt{ac}$ (assuming $b$ is positive, that is); however, this is not the same as $(b-2)\sqrt{ac}$! The parentheses definitely matter.
These are both the same "species" of error - they both involve misunderstanding how the various algebraic operations interact with each other. You can't rearrange operations willy-nilly: e.g. "adding, then squaring" is very different from "squaring, then adding", and so on.
There is a small simplification that can be made. Let's rewrite the quadratic equation as
$$ x^2 + 2B x+C=0. $$
Then the quadratic formula reduces to
$$ x= -B \pm \sqrt{B^2 - C}, $$
which is somewhat more palatable. It's occasionally more convenient in physics.