This is utility function that we will need to use in the next patch to
see how data of two blocks are similar to each other.

We use numpy for the implementation because this code will be hot and if we
don't use optimized C routines writeout will become very slow.

Quoting draft patch kirr/wendelin.core@3f631932 :

    -> Also optimize ndelta computation - when done in plain python just
       this part was taking a lot of time as timing for initial writeup
       showed:

         writeup with ZBlk0: ~20-25s
         writeup with ZBlk1: ~20-30s
         writeup with auto:  was ~ 120s

       now, after switching to numpy for ndelta computation, whole runtime
       with 'auto' is taking ~ 35s. The whole runtime, if I observe
       benchmark execution correctly, is dominated by database writeup.

/reviewed-by @levin.zimmermann
/reviewed-on !20