2nd draft stop metrics report

[tiffanychu90]

• do a 2nd draft portfolio following Add more stop time metrics data-infra#4308
• create sections for each metric and change charts
  • daily charts should be selectable in legend
  • show the day_type aggregation that is the desired grain to move towards for future charts
• add exploratory notebooks 07_stop_prediction_error_exploration and 08_stop_pct_accuracy to make sure the direction of signs matches interpretation
  • tag the 1% outliers of each tail and drop
• Use staging dbt models. Since they look good here, they will be merged in next batch of dbt models.
• Use google.cloud.bigquery to download warehouse tables, so array types get coerced correctly as arrays instead of strings.
  • Compared our existing sqlalchemy, pandas_gbq and google.cloud.bigquery and google.cloud.bigquery is at least 2x faster and it gets arrays out correctly
  • TODO: test out this hypothesis on vp path, the ultimate array + coercing string into geometries, and exporting as gdf
• 1st draft portfolio here: Plot RT stop metrics #1680

[github-actions]

nbviewer URLs for impacted notebooks:

• rt_predictions/07_stop_prediction_error_exploration.ipynb
• rt_predictions/08_stop_pct_accuracy.ipynb
• rt_predictions/rt_trip_updates_report.ipynb