This release has seen major revision to the bootstrap error analysis. I've known for a while that the way I computed the uncertainty in the free energy was suboptimal, since it just extrapolated from the uncertainty in the probability using the assumption that the fluctuations were gaussianly distributed. I knew this wasn't a great assumption, but Michael Shirts showed me data showing just how bad it was. So, in this version, we're doing something different. For 1D wham, we're directly computing the fluctuations in the PMF at each bin, aligning the pmfs such that their partition functions are all 1; this amounts to shifting them so that their Boltzmann-weighted free energies are the same. I think the errors make more sense now.
For 2D wham, I think doing this analysis uncovered some other flaws in how the error analysis is done, ones that to be honest I'm not totally sure how to fix. For now, I'm simply removing the option to do bootstrapping in 2D. I hope to put it back once I solve the problem, in which case I'll make another release.
Many thanks to Michael Shirts, for helpful discussions and contributing much of the code changes that went into this release.