Return sorted lists by exact_mappings

[nonhermitian]

Return mappings as lists that are sorted for use by initial_layout

Also had to bump call_limit to 1000 as otherwise VF2 was always missing a subgraph on the 27Q layout

[mtreinish]

On the call limit default value it makes sense to use a much larger number. In terra we use it roughly as a per iteration time limit, but it really is internal state visits in the vf2 algorithm and for larger graphs it will require a higher number to find any solution. In Qiskit/qiskit#7213 we used 50,000 for level 1 (which took about 100ms in local testing), 5,000,000 for level 2 (which was roughly 10 sec) , and 30,000,000 for level 3 (which was roughly 60 seconds). I have some benchmarks using vf2layour to solve for a perfect initial layout of a several hundred qubit circuit on a 1081 quit heavy hex coupling map and it taking about 20 seconds on average for the pass to find a solution and required the higher call limit to find a solution at all.

[nonhermitian]

Ok yeah that makes sense. There is probably a rough heuristic that one can find based on topology size and circuit size that gives a good guess for how many calls it takes. That of course assumes think like we always use heavy hex