In order to do that you will need to override the important internal routine toolbox/matlab/lang/@double/display to implement an additional "format" that has exactly three decimal places and which truncates. I would expect that would be a notable amount of work to get right without interfering with any of the current uses of the routine.
The built in "format" can display 2 digits after the decimal place ("format bank" or sometimes "format short g") or 4 digits after the decimal place ("format short" or "format short eng"), or 15 digits after the decimal place ("format long"), or up to 15 digits after the decimal place, stopping the remaining digits of the 15 are all 0. The "format" routines all round.
For any other format like 3 digits exactly with truncation, you need to override the display method. Or you could do the display formatting yourself, using sprintf() or fprintf()