Better diagnostics from solver

[sajith]

There should be less "magic" in TESolver.solve() -- we should be able to know why solver solved connection requests in a certain way, or why it was not able to find a solution, without too much effort and maybe by looking at some log messages.

Not sure how best to do this at the moment, since it can get pretty complicated when we have a large topology and a large connection request, for example. It would be useful to give this some thought.