Measure autocorrelation time of wordline `supercurrents'

[alcherman]

For large lattices in the worldline formulation, it seems that `holonomy' updates are accepted very rarely (acceptances rates of 10^{-3} for N ~ 24 and kappa ~ 0.8). This suggests that the associated autocorrelation time is at least 10^3, and much longer than any other correlation time. For example the autocorrelation time of the internal energy is about 3 for kappa ~ .8 . We should measure the largest autocorrelation time, and if it is indeed huge, consider switching algorithms to e.g. a worm algorithm.

[evanberkowitz]

In the Worldline case we know how to measure the supercurrents, it's just the particle flux across the boundary (which is also calculated in the TorusWinding in #58). But how do we formulate the equivalent thing in the Villain?

[evanberkowitz]

This is indeed markedly improved with the worm, which creates torus-wrapping updates at exactly the rate the configurations like!