Memory Leak; Old Solver Data Not Properly Deleted

[gunnarmiller]

There is a known memory leak issue in which previous sample sets recorded by the D-Wave solver object are not getting deleted in a timely fashion. These are moderately large memory allocations, at least several tens of megabytes, and each QPU call seems to add a new one, up to some limit after which memory use tops out. Including explicit "delete" statements to attempt to free up the allocated memory did not appear to help, so it seems the bug must be addressed at a lower level in the code for the solvers.

For a single solver, the total memory overhead of this leak seems to top out at slightly under 1 GB, meaning the code is memory inefficient, but still runs fine on a computer with reasonable resources. However, the bug can become more serious of multiple separate solver objects are instantiated: each one seems to separately use memory up to the limit. At present, the simulation should never have more than three solver objects instantiated at once, so the maximum memory overhead should still be bounded at about 3 GB.