As a pendulum gets longer the time period gets longer, so it is clear why T should be going up at the ends. When the bar is attached at its middle at the center of gravity, there is no change in gravitational potential energy as the bar swings, so there is no reason to oscillate, so T goes to infinity. Near there, with very little net torque to make it oscillate T is very large, so it is clear why it should go up in the middle. There must be a minimum in between.
Another way to look at it is comparing the torque to the rotational inertia. The angular acceleration is inversely related to the period, and angular acceleration is torque / inertia. So you can see how the period changes by examining torque / inertia. Picture the bar tilted from vertical by some particular angle. The torque increases linearly as you move the suspension point away from the center of gravity. However, the inertia does not change linearly. If L is the total length of the rod and x is how far the suspension point is from the center, then the moment of inertia is
## \frac 1 3 \frac M L (\frac {L^3} 4 + 3 L x^2)##
Which is parabolic in x.