rivers pack kind: OSM waterways lines for the moving map (W2 data side)

[billmallard]

Data side of billmallard/pyEfis#92 (moving map rivers-as-lines).

• New pack kind rivers (BULK, non-cyclical edition, like water/highways). MERGE THE KIND TO MAIN BEFORE FIRST R2 PUBLISH (2026-07-05 lesson: main's cron crashes on unknown manifest kinds).
• Builder: clone the roads build (work/build_na_roads.py) for gis_osm_waterways_free_1, keep fclass river+canal, drop stream/drain/ditch. Same sqlite schema as highways so pyEfis HighwayDB reads it unchanged. Douglas-Peucker simplify per the water/roads precedent.
• Updater: SQLITE_KINDS rivers -> rivers/rivers.sqlite; site + data.yaml.sample entries; picker group under Bulk.
• Size estimate: roads-CONUS was 108 MB / 851k polylines; river+canal should land under that (~40-80 MB, verify at build). California gotcha applies (2018 shapefile pin).

[billmallard]

Status: registration slice landed on dev, not closing yet.

31985c4 (dev) registers rivers as a first-class pack kind end-to-end in the data pipeline per this issue: packmeta.KINDS; core.py SQLITE_KINDS (rivers/rivers.sqlite), BULK_KINDS (region opt-in, like water/highways), KIND_LABELS ('Rivers & Waterways'); and a green test_rivers_install mirroring test_highways_install. The builder is the local work/build_na_rivers.py (gitignored, like build_na_roads.py) -- it extracts gis_osm_waterways_free_1 and reuses the highway_lines schema via build_highway_db.py --preset rivers, so pyEfis reads it unchanged.

Deliberately NOT done yet (house rule: land the kind + merge to main before the first R2 publish, or main's cron crashes on an unknown manifest kind): building/publishing a real rivers pack, the promote dev -> qa -> main, and the download-site + data.yaml.sample picker entries. Wiring the installed river_db_path into the moving-map instrument is separate map-provisioning, not the SVS options block (rivers has no SVS collector). Stays open until the pack is published and the picker/path land.