Add a batch callback version to p4est_search_reorder

[cburstedde]

Consider a batch-oriented callbock in search

The t8code folks report massive speedups in search when the callback is not executed per point but with an array that contains all currently active points. It would not be difficult to add p4est_search_batch as an extension of p4est_search_all, using the latest code from search_reorder, partition_gfp (note to self: progress #196) and a new callback design. We might still provide a Boolean parameter to the function to execute it with one point at a time, using the same call convention but with a 1-element array each time.

Opinions?