suggestion to merge rnn_new and rnn_old in kgraph.c

[gchen2101]

rnn_new is a list of objects whose distance to nhood_o is larger than nhood_o.radiusM. Each of rnn_new is an object taking nhood_o as a member of pool.
rnn_old has the same meaning but nn.flag is FALSE, which has no difference to nhood_o.

I would suggest using one list 'rnn' to replace rnn_new and rnn_old.
In the latter part, insert rnn to nn_old because rnn is not so important to objects.
Thus way, I think, can reduce the workload of join().