Output metric improvements

This includes two functional changes to how we compute rollout results and a few usability changes for the output metric dataframes.

Functional changes:

We now explicitly track year-by-year stock by asset type and vintage (year of investment), rather than only tracking fleet age in aggregate. This will improve precision of rollout investment schedules and resulting costs, and will handle some edge cases that were not accounted for previously.
Previously LCOTKM captured operating costs + investment costs - residual asset value at end of rollout. Now, it captures operating costs + investment costs - change in asset portfolio value. In other words, the calculation is the same except that starting portfolio value in year 0 is subtracted out. This is more in line with how fleet managers may track balance sheets, and it will make results more comparable across shorter-term vs. longer-term rollouts.

Usability changes:

Units for each metric are more human-friendly (e.g., ready for printing to plot labels as-is)
For all metrics that are either discounted or levelized, that is indicated parenthetically on the units to reduce potential for confusion
For stock counts, instead of metric and units both being the asset type (e.g., Metric="Count_Refuelers" and Units="Refuelers"), Units="assets". This is to make it easier to filter results to all asset investments regardless of asset type.

NREL / altrios