Align speed scripts part 2 / Jul open data part 2

[tiffanychu90]

segment_speeds

• Test on Jul 2024 for all segments, run all scripts from nearest_vp_to_stop to average_segment/summary_speeds
• Backfill all dates in rt_dates.y2024_dates, rt_dates.y2023_dates, rt_dates.apr2024_week, rt_dates.oct2023_week, rt_dates.apr2023_week...35 dates in total
• within pipeline_* scripts, there's a spot where kernel dies
  • start profiling the code again
  • wrapping all the dates in dask.delayed and then computing them seems to be ok.
  • only 1 spot where kernel will need to be restarted. after running nearest_vp_to_stop and after the 1st date in filter_to_two_vp. restarting allows the rest to run.
• multi-day averages need to be injected 1 week at a time, or kernel will crash
• what's probably happening is some X number of objects aren't getting trashed before the next date is computed, and at some point, some of the larger files overflow. restarting trashes everything and subsequent steps are small enough where this isn't an issue.
• Research Request - Align segment_speeds pipeline scripts to use vp with dwell time #1183
• run Jul 2024 for Epic - Open Data Publishing 2024 #991

gtfs_funnel

• Update README and add a table to reflect what dates are supported across these workstreams

rt_vs_schedule

• overwrite and backfill rt_stop_times tables for all dates
• rest of rt_vs_schedule metrics does not change -- use vp_usable without the dwell times, because we want to know how many vp are produced each minute, and we don't want a vp that's been rolled up across minutes
• run jul 2024 for monthly publishing