tools.costs outputs all combinations of year_vintage and year_act for technologies

[measrainsey]

As highlighted in the comments in PR #222, the final dataframes created by the costs tool has every combination of year_vintage and year_act for technologies. The tool does not take into consideration the lifetime of the technology, so there are projected costs for year where a technology is no longer active.

This doesn't affect the solution in MESSAGE, but it is problematic for a number of reasons:

• The output dataframes are bloated.
• It could be confusing for users.

Some possible existing tools/functions that can help address this:

• message_ix.vintage_and_active_years()
• message_ix_models.model.water.util.map_yv_ya_lt()
• message_ix_models.util.ScenarioInfo.yv_ya

I didn't want to implement this fix in an existing PR, in case it gets too big, so creating this issue to keep track of this as part of my TODO.