One-norm does not account for double factorization truncation threshold

[max-radin]

Right now, resource estimates for DF QPE might be somewhat overestimated because the one-norm calculation in pyLIQTR.ProblemInstances.ChemicalHamiltonian.get_alpha does not account for the double-factorization threshold df_error_threshold. (Note: I've already discussed this with @kjm538, just posting an issue here so that it's easier to keep track.)