We've been experimenting with this ourselves, having recently moved to CPU with this architecture.
The NCrunch settings wizard is ignorant of the characteristics of the cores, so it's probably best to set this manually in the configuration window.
Assuming you're running VS2022, I would recommend giving this a healthy share of the performance cores, since most of the work this process does is very time sensitive.
In theory, as long as you have at least one performance core assigned to NCrunch, the O/S should give it preference for any work that NCrunch does. This would allow you to get fast returns on tests executed with priority with an otherwise idle engine. In this situation I am making an assumption that the O/S would favour the PCores over the ECores (which we would all expect it to).