CHANGELOG

#######################################
#           Aboria 0.5                #
#######################################

Summary

    This version has extensive changes to the backend iterators that are used to interact
    with the neighbourhood search data structures, to fit in the new tree-based structures.
    In particular, there is a new `child_iterator` that is used to iterate through children
    of a node in a tree, a depth-first search iterator, and a query iterator specificly
    for neighbour searches on a tree
    
    There are also large changes with the backend neighbourhood search functionality. Now 
    a distance search can:
        * be anisotropic, the distance measure is scaled independently in each dimension
        * use any L-norm for a distance measure, L1, L2, L3, ...., Linf
        * can handle periodic domains independently of the data structure
        * can handle any radius, independently of how the data structure was initialized 
          (see below)

    Finally, the syntax for the `Aboria::Particles::init_neighbour_search` does not require
    the user to enter in a cut-off radius, instead the user can enter in a number corresponding
    to the average (for cell lists) or maximum (for trees) number of particles in each cell/bucket.

    New search data structures! Finally have a kd-tree (courtesy of nannoflann 
    - https://github.com/jlblancoc/nanoflann), and an octtree. These are used for the new 
    fast multipole method and hierarchical matrix kernel operators.

    New kernel operators - Aboria now has implementations of the black-box fast multipole
    method, and the ability to store this as a H2 hierarchical matrix for repeated evaluations.
    Please see documentation for more details. btw. The previous functionality termed "matrix-free"
    operators has been renamed to "kernel operators" to factor in these changes.