improve solution for R-tree key error

[rcgorman-dstl]

generate_hypotheses method in TPRTreeMixIn class in tree.py file was previously throwing key errors when finding intersections in some loop iterations. Try/except block was added to reset TPR tree when key error is encountered and rebuild the TPR tree from scratch as a workaround to run successfully. As key errors are frequent, this is likely increasing run time of data association using TPR tree and resulting in sub-optimal performance. It is not clear why key errors are thrown in the first place so this requires further investigation so a better fix can be provided.