Support remap for adaptive refinement with incrementally appended source/target sites

[kuberry]

• Ability to add a single or set of source sites, using these sites with a neighbor search among target sites to find which target sites need their neighbor list updated (then update the neighbor lists).
• Add ability to recompute solution for all target sites affected by the source site appending.
• Similar functionality for dropping source sites, equivalent to coarsening.

[kuberry]

• Also consider a reorder function that takes a map that is the size of the current target sites. It can then migrate any solution data, target coordinates, and neighbor lists, from the previous state to the reordered state (with the ability to have a reordered state that has more or less target sites than existed before reordering).
• Could be a 2D array so that the first column could be 0 or 1 (whether reordering changes the value), and the second column is the designated new index after reordering. This would permit creating a 2D array of zeros and then only modifying entries desired.
• Batch removal of target sites could be written as a helper that calls the above function and creates this 2D array for removing target sites.

[kuberry]

NeighborLists class is now a compressed row format.

• While alpha calculation is in GMLS class, it may make sense to extract the solution and any relevant information needed to do a reconstruction into a GMLS solution class.
• Then, if a single additional site needs a reconstruction, the solution can be stored separately.
• GMLS solution merges could work like a std::vector, allocating extra space (if desired) so that appending one extra solution could fit without reallocation.