You probably don't want to change maxsearchesper_cpu in limits.conf because it would be a global change and could have a detrimental impact to performance.
I think you're asking how to bypass the concurrent search limit for a user or a role, right? You can create a new role in Splunk Web (under Access Controls), set the "Role-level concurrent search jobs limit" to 100, "User-level concurrent search jobs limit" to 100, save the role, and then add that user to the new role you created.
Note: If you are maxing out cores, it's probably time to add indexers (so that searches complete faster), or add another search head if you have a lot of scheduled searches running all the time.
The "data import" user has already an admin role, so its limits are much higher. I suppose
max_searches_per_cpu has a much higher priority, and I'm searching a way to bypass it.
Yes, adding search headers is a good workaround, but it's not an optimal solution. This way is very simple to do a DDoS attack that will prevent data importing.