1. The most significant property of a diode connected transistor is that it is always in saturation (Vds = Vgs => Vds > Vgs - Vth).
Now consider the behavior of this device. If the drain potential is higher than the source, it will be in saturation and the current will flow from drain to source, which justifies the direction of the diode in the picture.
A MOS transistor is generally symmetrical (either terminal can be drain or source). For an NMOS the terminal at the higher potential is designated drain and that at the lower potential is called source. Therefore, if the polarity of the voltage across the MOS is reversed (the terminal marked drain in the figure behaves as 'source' and vice versa), there will be no current (Vgs=0).
This explains the behavior as a diode. It only allows current to flow from drain to source if the drain potential is higher than the source potential.
2. Again, the diode load is not really an ideal diode but actually a MOS device biased always in saturation. The equivalent resistance of this can be computed easily using an AC model or you could reference any textbook ( Design of Analog CMOS ICsj by Razavi, Ch 3). For a diode load Rds ≈ 1/gm.
3. Depletion devices can be thought of as negative threshold devices. They can conduct current for zero Vgs. So in this case, the direction of the diode will be reversed (Based on the preceding logic, do you understand why?)