Fix outlet ordering + extract reusable trace

[NewGraphEnvironment]

Summary

• Fix lake outlet ordering: use wscode_ltree + localcode_ltree (network topology) instead of downstream_route_measure (meaningless across BLKs) for multi-BLK waterbodies
• Fix cumulative distance partitioning: partition by waterbody_key instead of blue_line_key so distance accumulates per lake trace, not per BLK
• Extract .frs_trace_downstream() — reusable downstream trace with distance cap + gradient stop, takes any origins SQL
• Rename .frs_connected_spawning() to .frs_connected_waterbody() — parameterized waterbody_poly table supports lakes and wetlands

BULK SK spawning: -22.6% → +0.1% vs bcfishpass.

Test plan

• 696 tests pass
• Code-check clean (2 rounds)
• Verify BULK SK spawning matches bcfishpass within tolerance
• Verify ADMS SK spawning unchanged (single-outlet lake unaffected)

Fixes #147
Relates to NewGraphEnvironment/sred-2025-2026#16

🤖 Generated with Claude Code

[NewGraphEnvironment]

Reservoir support needed in Phase 2

bcfishpass load_habitat_linear_sk.sql lines 217-240 checks BOTH lakes AND reservoirs in the upstream cluster proximity:

clusters_near_rearing as (
  select c.cid, bool_or(lk.waterbody_key IS NOT NULL) as wb
  from clusters c
  left outer join whse_basemapping.fwa_lakes_poly lk
    on st_dwithin(c.geom, lk.geom, 2)
  WHERE lk.area_ha >= t.rear_lake_ha_min
  group by c.cid
  union all
  select c.cid, bool_or(res.waterbody_key IS NOT NULL) as wb
  from clusters c
  left outer join whse_basemapping.fwa_manmade_waterbodies_poly res
    on st_dwithin(c.geom, res.geom, 2)
  WHERE res.area_ha >= t.rear_lake_ha_min
  group by c.cid
)

The current PR accepts a single waterbody_poly table. This misses reservoirs. Options:

1. Accept a vector of waterbody tables: waterbody_poly = c("fwa_lakes_poly", "fwa_manmade_waterbodies_poly")
2. Build the UNION inside .frs_connected_waterbody() when multiple tables provided
3. Pre-union the polygons into a temp table before calling the function

Also: Phase 1 (downstream trace) identifies lake outlets from rearing segments. Rearing on reservoirs also needs outlets traced downstream. The lake_outlets CTE filters by hr.rearing IS TRUE which includes reservoir rearing, so Phase 1 should be OK. But Phase 2's ST_DWithin against a single poly table misses reservoir clusters.