I have a computation over the thousands of items that needs to process a sequence of computation where each n iteration need to process data of [n - 1, n, n + 1]. Better to understand this on the chart 
All the computations inside iterator are same so we can cache it. But what is the best way to do it with rayon?
I thought to cache it based on the thread level. So if we have 4 threads that iterates over we can have cache for each group (0 - n/4, n/4 - 2*n/4, ..) but we can not control the sequence of iteration in rayon.
Is there any better way of having per-thread or shared cache in rayon without mutexes?
Important note: I don't know which iteration will need to make a calculation. It determines in runtime, so basically I can not make all the calculations before iterations.
