Convex approximation of charge tiers leads to negative shadow costs

[arao53]

Issue:
Right now, the charge tier is fixed at the beginning of the optimization problem, if the optimal solution violates this assumption then the actual cost (evaluated in post) may be different (usually higher). This can result in issues in calculating the shadow cost of emissions abatement, especially if the calculated cost-optimal cost is higher than the emissions-optimal cost.

Potential fix:
Two options:

1. iteratively solve tariffs and re-build the costing equations by re-calculating the consumption dict. After a few iterations, this should make post-optimization cost evaluation = within optimization cost variable. No guarantees.
2. implement charge tier constraints within the optimization problem (in electric-emission-cost).

[fletchapin]

PR out for a bug fix in EEC package: we3lab/eeco#35

Please test on the hotfix/charge-tiers branch before merging.