Add interpolation p2

[tiffanychu90]

segment speeds

• after moving the direction check to all shapes (not just loop_or_inlining==1 shapes), finish up rest of workflow to get speeds
  • check that low speeds were happening also on loop_or_inlining==0 shapes. we can't ignore these.
• add a step to re-calculate speeds if pct_segment_covered < 30% (if it has 1 point, the segment would have had zero meters_elapsed, because the same point would have been chosen for first and second point within segment).
  • for these points, choose the previous vp_idx. if we're close to the origin and there's no previous point, choose the following vp_idx, and use absolute value for the distance and time calculation.
  • chose to use straight distance because the prior_vp_idx could take place on a separate segment.
    • if we use shapely.project and hit the entire shape geometry, we risk getting erroneous shape_meters calculated. no way to know if it's projecting onto the inbound vs outbound portion.
    • decide that straight distance will be close enough, if points are close together (yes, these are consecutive vps), while handling the fact that they may not fall on consecutive segments
• add a flag_recalculated column to track whether the speed was calculated again with straight distance
• adjust avg_speeds script to remove the pct_segment argument. it now appears in an earlier step.
• future TODO: should we add a map_partitions way of calculating? test a full run in oct to see how long it's taking, now that we swapped some steps, see how it's shaking out
• future TODO: now that direction check occurs earlier, we can apply this workflow to road segments, which would not have a loop_or_inlining flag