Colorful Traveling Salesman Problem datasets

=== File format
Each line in the files has the format A,B,C where A is the start vertex of the
specified edge, B is the end vertex of the specified edge, and C is the label
associated with that edge, with A < B (because graphs are symmetric for the
CTSP). Both vertex indices and labels are 1-indexed (aka numbering starts
from 1).

=== Files
The files l_*_*.gml are from the following paper:
N. Jozefowiez, G. Laporte, and F. Semet. A branch-and-cut algorithm for the
minimum labeling hamiltonian cycle problem and two variants. Computers and
Operations Research, 38(11):1534--1542, 2011.

The first number in the filename is the number of nodes and the second number
is the number of edges in the problem instance. Each of these numbers is in
the set {50, 100, 150, 200}, leading to 16 problem instances.

The remaining files are based off of four TSPLib problem instances: eil101,
a280, att532, and dsj1000. The CTSP datafiles are called, respectively,
*_*_102eil101.gml, *_*_784a280.gml, *_*_2830att532.gml, and
*_*_10000dsj1000.gml. In the naming convention, 102eil101 means the dataset has
102 edges and 101 nodes. The dataset generation approach is from the paper:
Y. Chen, N. Cornick, A. Hall, R. Sahajpal, J. Silberholz, I. Yahav, and B.
Golden. Comparison of heuristics for solving the GMLST problem. In S. Raghavan,
B. Golden, and E. Wasil, editors, Telecommunications Modeling, Policy, and
Technology, pages 191--217. Springer, 2008.

The first name in each filename is the generation format (one of LC, LR, RC,
RR, as described in that paper), and the density (one of 0.2, 0.5, 0.8, or
1.0). Hence, there are a total of 16*4=64 TSPLib-based instances.
