Backfill historical data through new analysis pipeline

[rviscomi]

After #19, the pipeline should be able to automatically process new monthly datasets. To make sure that the trends on the HA website are useful, we would need to backfill historical data through the pipeline.

By default, the analysis pipeline should automatically run on the latest monthly dataset. We should also be able to manually run the pipeline on historical datasets to generate new metrics from old data.

Once that functionality is set up, we should run all applicable historical datasets through the pipeline to generate the historical trends for each metric in the HA reports.