It will be in Options, Advanced.
Which ones does it start with?
High, and Maximum.

That's a bit extreme, just saying.
Generating a crappy motion blur would probably work, it would probably look worse then having no motion blur and it'd make the game much slower.
Generating good motion blur that looks nice, while doing that real-time takes serious power. Even with the kind of hardware you'd need for that it'd still probably be laggy.

Rendering a 10s clip with moderate editing and effects would probably take about 20 minutes tops. Adding RSMB (motion blur) to that can triple the amount of time needed depending on a bunch of factors. That's just some perspective on how much power it takes to render that, and that's not even real-time. Doing those calculations real-time is just unthinkable and I doubt that anyone in the community has the kind of hardware required to do something like that.

Not to shut down your idea; I think it'd be awesome; but it's probably just not feasible to do that real-time. It's a bit ridiculous.
This wouldn't be under maximum; it'd be under insane.

Realtime is quicker than video stuff, that's why it's such a common effect in games.

If we were able to make full screen fragment shaders for TB it could be done easily, and the hardware requirement would be negligible. We're talking old bottom end DX8 hardware being able to do it without breaking a sweat... Well lets just say you could do it realtime on hardware that isn't even capable of running TB properly without any noticeable FPS hit lol.
It's only a 2 operation effect to pull off, your toaster could probably do it.

it used to be a thing but looked pretty bad

The old one was a mess, it was a DOF effect being applied selectively to anything that was moving over a certain speed, hence it looked like mud and had a weird cut off.
We had a full DOF command back then too, but no one used it because it raped FPS & (for the time) had hardware requirements most players couldn't keep up with.

People tend to over complicate the effect & make effects that guess what should be happening, but it looks wonky and can be pretty heavy on hardware. Simply buffering the previous frame & using it to create a blurred after image is all that's needed - exactly what video editing stuff does.
(except video effects have to calculate everything to figure out what's moved between frames hence it's slow, not something we'd have to do)
It's perfectly doable in realtime.