Jaycanwk, your question is one that was asked in the very early days of decompression science, back in early 1900's. At that time, "caisson disease" was a big problem for construction crews (aka "sandhogs") working in a pressurized caissons beneath rivers, doing things like building the foundations for large bridges. Caisson disease is what is now called the bends or DCS.
One researcher, Hill, proposed a linear ascent, such as you mentioned in the opening post. J.S. Haldane proposed a staged decompression approach, which spends more time shallow than deep. Haldane's approach proved to be much better.
https://diversalertnetwork.com/membership/alert-diver/article.asp?ArticleID=573
While there is some difference between various models in the exact shape of the ascent depth vs. time curve, all models spend more time shallow than deep.
One could efficiently offgas using a constant ascent, but the speed of ascent would be faster when deep; slower when shallow. On a practical basis, experience has shown that a stepped approximation of this ideal NON-linear ascent by a series of stops 10' apart is close enough that going to a continous (but varying with depth) ascent doesn't improve the decompression enough to make it worth the extra effort as compared to 10'/3m stops.
-------------------------
Decompression is a balancing act. One wants to get shallow in order to stop further ongassing and to promote offgassing. OTOH, one wants to stay deep in order to prevent bubble formation. These are conflicting goals, and the "faster when deep, slower when shallow" type of ascent has been found to be the best balance between these conflicting goals.
Bubble models such as RGBM and VPM differ slightly from neo-Haldanian models in that the bubble models spend a bit more time deep than does the Haldane model, but the basic ascent curve for both is weighted with more time shallow.