GoingToCamp provider & Washington State Parks

User stories

• Campers discover, map, and monitor Washington State Parks campgrounds and overnight lodging (https://washington.goingtocamp.com) through the existing application.
• Operators sync complete park and campground metadata independently of availability scans via CLI or scheduled background ingestion.
• Maintainers add future GoingToCamp park systems (Parks Canada, BC Parks, Wisconsin, Maryland, Michigan, etc.) as thin deployment variations on a shared async ABC rather than duplicating adapters or hardcoding fake recreation-area IDs.

Requirements

• Implement a shared async base GoingToCampProvider(BaseProvider, ABC) in camply.providers.going_to_camp.provider with source Pydantic v2 models in camply.providers.going_to_camp.models.
• Model each GoingToCamp website as a concrete provider variation with its own persistent Provider identity because upstream transactionLocationId and resourceLocationId sequences independently start at -2147483648 across deployments and would otherwise collide on (id, provider_id) primary keys.
• Register WashingtonStateParksProvider in camply.providers.going_to_camp.variations under permanent provider ID 12 (Washington State Parks, https://washington.goingtocamp.com).
• Seed provider ID 12 through an immutable Alembic migration revising the current head, and block downgrade when dependent rows exist in campgrounds, recreation_areas, unique_targets, or search.
• Ingest complete park (GET /api/transactionlocation) and campground (GET /api/resourceLocation) catalogs independently of availability scans, resolving active overnight camping/lodging categories dynamically from GET /api/bookingcategories and GET /api/resourcecategory instead of upstream's hardcoded negative category IDs.
• Retain day-use, marina/moorage, disabled, and unmapped facilities in campgrounds with reservable=False while excluding them from searchable Campground entries in search, and retire disappeared upstream records (enabled=False, reservable=False) without deleting historical scan references.
• Scan campsite availability asynchronously in bounded date windows (_GRID_DAYS = 60) using bulk facility resource metadata (GET /api/resourcelocation/resources) and map daily availability (GET /api/availability/map?getDailyAvailability=true&isReserving=true), traversing child loop maps via mapLinkAvailabilities only when potentially available (0 or 7) and treating only status 0 (Available) as bookable online.
• Treat end_date as check-out (exclusive), clip padded or check-out daily entries to [start_date, end_date), merge multi-window results per resourceId, and raise ValueError when an unknown campground ID is requested so cached worker scan results are preserved.
• Decode equipment across all equipmentCategories from GET /api/equipment and filterable multi-enum/scalar site attributes from GET /api/attribute/filterable, conservatively classifying CampsiteType (CABIN, RV, TENT, OTHER), is_electric (excluding explicit non-electric values), and is_accessible (restricting matches to accessibility attributes rather than adjacent amenities).
• Generate verified GoingToCamp reservation deep links (/create-booking/results?transactionLocationId=... and /create-booking/results?resourceLocationId=...) for parks and campgrounds.
• Keep all committed fixtures and VCR cassettes synthetic, with behavior regressions covering HTTP replay, map traversal, category/equipment/attribute decoding, metadata atomicity, provider isolation, CLI help/invocation, packaging, and migration round trips.

Scope limits

Remaining GoingToCamp deployments from the legacy CLI are documented in provider TODOs for per-deployment endpoint verification and rollout. Individual campsite catalog persistence, cross-worker distributed rate limiting, and shared-worker minimum/maximum stay enforcement remain platform-wide follow-ups. No legacy CLI code is imported and no new third-party dependencies or database schema changes are introduced.