How can a gas be in circular orbit when centrifugal force is dominant in an accretion disk around a black hole? In the textbook "Black Holes, White Dwarfs and Neutron Stars: The Physics of Compact Objects" by Shapiro and Teukolsky, the author makes the following statement while discussing the standard thin disk model:

Whenever the angular momentum per unit mass, $\tilde{l}$, exceeds $~r_Ic$, where $r_I$ is the innermost stable circular orbit, centrifugal forces will become significant before the gas plunges through the event horizon. In this case, the gas will be thrown into circular orbits about the black hole, moving inward only after viscous stresses in the gas have transported away the excess angular momentum. 

I am having the trouble to understand how the gas will be thrown into circular orbit when the centrifugal force is significant, since centrifugal force is always directed outward from the axis of the rotation.
 A: Most of the physics that's relevant here isn't qualitatively different from what's involved in the Newtonian limit of weak gravity. A particle in a Newtonian orbit won't approach closer than a certain perihelion distance $r_0$.
In a non-rotating frame, one way of describing this is that if $r$ is to get small, then the particle's velocity has to get big in order to conserve angular momentum. Although conservation of energy does cause the velocity to get bigger as $r$ gets smaller, the rate of increase isn't fast enough to allow $r<r_0$.
A whole different mode of description is to adopt a rotating frame, in which the particle is at rest when it reaches $r_0$. An observer in this Newtonian, non-inertial frame sees the particle curve inward, pause at $r_0$, and then accelerate back out. In this frame, we have a fictitious centrifugal force, which is what repels the particle back outward.

I am having the trouble to understand how the gas will be thrown into circular orbit when the centrifugal force is significant, since centrifugal force is always directed outward from the axis of the rotation.

They're trying to explain intuitively why the particle doesn't keep falling inward and pass through the horizon.
For the actual quantitative analysis of the orbits, people don't actually use these modes of reasoning. Typically one reduces the motion to a single variable $r$. When we do this, we get something that acts like a potential, and the shape of the potential depends on the angular momentum.
