Leaving a lateral distance of one width to each boundary may be good; and, of course, you need also an appropriate representative depth for your model that depends on the subsoil layering and on the size (height, weight) of your structure under and above the ground.

Anyway, if you do not have a big limitation of computational power, the best thing you can do is to make some preliminary sensitivity analyses: see how the result changes if you leave a distance of 0.5 widths, 1 width, 2 widths and so on, so you can be sure to choose a value that lets you exclude significant border effects.

Commonly we use twice to three times the larger dimension of the structure from each side. I do not have a reference but it is just the common practice.

check the attached paragraph from "Geo technical modeling", David Muir Wood-2004

Thank you guys ever so much! I am grateful to you if you could share some high-ranked references wherein some practical suggestions were put forth.