How many depends on a lot of things ranging from how much you know about reloading and your chosen powder/bullet/primer/brass through how and what you plan to test.
For a first-time firearm and a first-time powder and with no chrono, I'd make up five rounds at each powder level and I span the entire range of possible loads. From the lowest load, I fire the first shot as a single and - if the brass et al look OK - the remaining four shots are then loaded and fired. If those four brass look OK, I move to the next higher level, rinse and repeat. Depending on what happens, I may choose to stop, not fire, and later disassemble the "hottest" load(s).
For accuracy, in a second session I'd then make up 10 of several of the loads that I have proven are OK to use. If the specific accuracy test pattern I'm running requires more, I load more. If using 5-shot groups as the measure, 10 rounds gives two test runs (if needed) at each powder level I'm testing. Anything left over after testing can be plinked since it is already proven "safe".
Chrono data, if available, helps to judge several things, from consistency to the safety of a given load . . . and the next higher one. In my case, the only chrono I have is for rifle only and since it attaches to the barrel, accuracy is distorted. So for the above "accuracy test" I might make up 15 rounds and use a random 5 to establish velocity.