If you plan to use the hammers for conquest then you may find that four is sufficient. Of course, more is better (although no, a unit can't use more than one hammer at a time) - but as you noted the longer you wait to cast the spell the longer you must go without the hammers. Rather than picking a specific number of cities or a specific turn limit, I would suggest casting the spell as you are getting ready for your first attack against a neighboring civ. If you end up with more hammers than you have attackers then you need to take a larger army.
If you plan to settle the hammers as engineers then I recommend about eight cities. Here, too, more is better, but you don't want to wait so long that the benefit comes when it is no longer necessary. Eight will give you decent returns and doesn't take too long to set up; I wouldn't push for more than that. If you run out of room to expand, though, don't hesitate to use the hammers before you have eight cities. Better to have that production to build your army than to wait until you can capture more cities.
If you are working toward summoning the Mercurians, then I suggest that you use the spell when either A) you are ready to start building the gate, or B) you know that you won't be building any more cities before you start building the gate. With this strategy you definitely want to have as many engineers in the gate city as possible, but you also want that gate to finish as quickly as possible. Don't hold off casting the spell while the gate is being built to try to squeeze out a couple more cities. The extra engineers are not worth the risk of being beaten to the gate, nor are they worth the lost turns resulting from a delayed completion of the gate.
