Well ...
What really matters is the rise/fall time of the buffers that will actually drive the capacitive load of the mixers switches. It is not a good idea to drive the mixers with ideal voltage waveforms with finite rise/fall times because ultimately the rise/fall times are limited by the driver slew rate and parasitic capacitance.

What are you trying to emulate by setting the rise/fall time?

An ideal voltage source will drive the gates without any consideration for the capacitance load of the gates; therefore, the rise/fall time of the voltage source (I assume you are driving the gates with a voltage source) is meaningless.

You need a buffer; otherwise, what are you trying to model by using the rise/fall time?

Ok, then I would sweep the rise/fall time to see the impact on mixer conversion gain/loss, linearity and NF.

The maximum rise/fall time should be fast enough for the pulse to transverse from 0 to the supply voltage and back down to 0 for the shortest pulse width you need (i.e. highest LO frequency).