Refactor codebase following Clean Code principles and add 229 tests

- Extract helpers to reduce function sizes (listing_tasks, app.py, query.py, listing_fetcher)
  - Replace nonlocal mutations with _PipelineState dataclass in listing_tasks
  - Fix bugs: isinstance→equality check in repository, verify_exp for OIDC tokens
  - Consolidate duplicate filter methods in listing_repository
  - Move hardcoded config to env vars with backward-compatible defaults
  - Simplify CLI decorator to auto-build QueryParameters
  - Add deprecation docstring to data_access.py
  - Test count: 158 → 387 (all passing)

crawler/services/floorplan_detector.py

@@ -6,12 +6,14 @@ from repositories.listing_repository import ListingRepository
 from tqdm.asyncio import tqdm
 import multiprocessing
 
+# Use a quarter of available CPUs to avoid starving other processes
+MAX_OCR_WORKERS = max(1, multiprocessing.cpu_count() // 4)
+
 
 async def detect_floorplan(repository: ListingRepository) -> None:
     """Detect square meters from floorplan images for all listings."""
     listings = await repository.get_listings()
-    cpu_count = multiprocessing.cpu_count() // 4
-    semaphore = asyncio.Semaphore(cpu_count)
+    semaphore = asyncio.Semaphore(MAX_OCR_WORKERS)
 
     updated_listings = [
         listing
@@ -29,6 +31,9 @@ async def _calculate_sqm_ocr(
     """Calculate square meters from floorplan images using OCR."""
     if listing.square_meters is not None:
         return None
+    if not listing.floorplan_image_paths:
+        listing.square_meters = 0
+        return listing
     sqms: list[float] = []
     for floorplan_path in listing.floorplan_image_paths:
         async with semaphore: