perf: most performant .return_value calculation

[antazoey]

What I did

Finally - .return_value is speedy as it ever was.

stats:

 Before:
 10 passed in 197.45s (0:03:17)
 After:
 10 passed in 69.65s (0:01:09) 

How I did it

if possible, use a barely enriched raw call tree, as opposed to a fully enriched (least performant) and a trace-calculated return (middle performance).
this is by far the best.

How to verify it

Checklist

• All changes are completed
• New test cases have been added
• Documentation has been updated