Starver, the actual implementation could be done any number of ways, but if you boil it down, basically I think the Bookkeper should have a job to do that is a continuous job in most forts. Keeping up with masses of stone.
As the simplest-possible implementation I can imagive, with little to no flavor at all, you could simply add a feature to each stone-using workshop so you could tick off the option for the bookkeeper to organize shortest path calculations for the nearest <x> rocks to a workshop (or rock stockpile). Something like designating what you want to grow in a farm.
You don't have to keep the pathing calculations in memory. Having precalculated them, if you direct the dwarves to get a stone from a certain tile, they will know to path to that tile when that specific rock comes up in the list. Dwarves already path relatively decently when they are told exactly what to do, it's when they choose for themselves with current algorithms that they will travel 74 z levels down, 47 south, 13 up, and 44 north to pick up a rock when there's a stockpile six tiles away.
In order to keep track of how fortress changes impact pathing, you might store one piece of additional data. Path length. Give location and path length for rocks, and when a dwarf consults the table and creates his path, if the path he creates is more than 20% longer than the path length recorded, the dwarf does a simple sort and puts the stone at the proper location in the lookup list based on the current path distance. Any dwarf should be able to scribble "New Wall. This rock now 44 distant, not 21" and move it down the list, then choose the next closest rock on the list. The bookkeeper would create the list, and periodically add to it, but every hauler can scribble edits onto it, which should probably be a no-gainer skill like assembling constructions or you would have a fortress full of bookkeeper haulers in a short while
If you set a stockpile as the basis for the quartermaster's calculations, then obviously that stockpile itself should be ignored for purposes of rock collection, or you would waste a LOT of CPU cycles calculating to move rocks around inside stockpiles, which would be a pain. Workshops should look to stockpiles as possible sources, passively. If the stockpile is closer than rocks, any quartermaster work would actually be calculated from the stockpile instead, by cascade, that way you don't build multiple redundant lists of data.
It might sound messy being described in detail, but in it's crudest form, it's really just a real-path-distance-sorted list, with the origin either at the point of use, or at the closest stockpile, if there is an appropriate stockpile closer than appropriate stone. I think adding a little flavor to it would be good, but it can operate in crude form too.
Only one dwarf to do the basic job of quartermastering. Parties are designed to interrupt fort operations and force you to create organically survivable forts
If the quartermaster isn't doing their job, the dwarves just revert to normal behavior.