I believe the order you're looking for is automatic.
I wanted to add that in my experience the file names don't matter either. It processes them by csv headers, specifically it recognizes them as instructure_csv. I send it multiple files of each type with each import, this leads me to believe it collects all of the files and sorts them and imports them by hierarchy.
My zip archives basically look like this:
| file | api |
|---|
| -employees.csv | users.csv |
| -employee-workrooms.csv | courses.csv |
| -workroom-enrollments.csv | enrollments.csv |
| -site-conference-enrollments.csv | enrollments.csv |
| -students.csv | users.csv |
| -sis-courses.csv | courses.csv |
| -sections.csv | sections.csv |
| -student-enrollments.csv | enrollments.csv |
| -teacher-enrollments.csv | enrollments.csv |
with batches and the file prefix being either weekly-, daily-, hourly-
and hourly files with 0 lines are removed before zipping
and the final import status will show "supplied_batches":["course","section","user","enrollment"] and the counts will have combined numbers
also https://community.canvaslms.com/thread/22550-peoplesoft-canvas-real-time-enrollment-sync#comment-998...
