Add a rank column to conflated.parquet for prioritizing downstream work #713

Summary

Add a rank column (or similar) to conflated.parquet, so downstream users can work through unmatched/conflated features in priority order instead of an arbitrary order.

This is a placeholder issue to track the idea. We'll flesh out a detailed design separately and replace this text with the actual plan once that's done.

Rough idea

Rank would likely be a blend of several signals, merged into one score:

• QRank (https://qrank.toolforge.org/) for features carrying brand:wikidata — a measure of how significant the linked Wikidata item is.
• OSMViews (https://osmviews.brawer.ch/) based on the centroid position — a proxy for how much attention/traffic a location gets; see https://github.com/brawer/osmviews for background. Rust client in https://github.com/brawer/osmviews-rs.
• A custom tag-based rank (to be designed) so that, e.g., at roughly the same location a fire hydrant should rank below a store, even when the store has no brand:wikidata tag to hang a QRank off of.
• Possibly OSM edit/version count (we already carry the version number through the pipeline) as a signal that frequently-edited features are more likely to be important.

Next steps

• Design how the individual signals get normalized and combined into a single rank.
• Decide where in the pipeline the rank gets computed and attached.
• Update this issue with the concrete plan before implementing.