Zygote `rrule(typeof(apply),...)` performance drop

[GTorlai]

Looks like the time-to-solution to obtain gradients from TEBD simulation of dynamics (i.e. OCC) have substantial overhead compare to simple evaluation of the function.

[mtfishman]

Could you share a benchmark code and some timings? What's the ratio of the time to compute the gradient to the time of just computing the function?

[GTorlai]

This performance drop has been reported by a user. I will work to produce a minimal example. I just opened this not to forget.