Durability pass: tests, clearer diversity/classify behavior, Calm Filters foundation

- Add pytest suite (34 tests) covering scoring thresholds, dedup clustering +
  representative selection + time window, brief source/category diversity,
  avoid-term phrase matching, and text canonicalization/truncation.
- Rewrite _select_diverse with an explicit, tested contract (best-first, one
  per source, backfill, then inject a second category by evicting the
  lowest-ranked pick).
- classify_articles now returns attempted/succeeded/skipped (ClassifyReport) so
  silent model failures are visible in both the cycle and classify output.
- Fix clean_text truncation to stay within max_len (ellipsis no longer
  overshoots).
- New filters.py: canonical FilterPrefs shape (include/mute topics+flavors,
  avoid_terms, pauses) and pure word/phrase-boundary matching engine seeding
  Calm Filters. Not yet wired into the API.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

goodnews/llm.py

@@ -220,6 +220,14 @@ class LocalModelClient:
         return parse_classifier_json(content)
 
 
+@dataclass
+class ClassifyReport:
+    results: list[tuple[int, dict]]
+    attempted: int
+    succeeded: int
+    skipped: int
+
+
 def classify_articles(
     conn: sqlite3.Connection,
     client: LocalModelClient,
@@ -228,17 +236,19 @@ def classify_articles(
     dry_run: bool = False,
     only_unclassified: bool = False,
     progress: "Callable[[int, int, int], None] | None" = None,
-) -> list[tuple[int, dict]]:
+) -> ClassifyReport:
     rows = _classification_candidates(
         conn, limit=limit, include_rejected=include_rejected, only_unclassified=only_unclassified
     )
     results = []
+    skipped = 0
     for index, row in enumerate(rows, start=1):
         try:
             scores = client.classify(row)
         except RuntimeError as exc:
             # One slow/failed article (timeout, bad response) shouldn't sink the
             # whole batch or discard work already committed. Skip and continue.
+            skipped += 1
             print(f"[{row['id']}] skipped: {exc}")
             continue
         scores = normalize_scores(scores, model_name=client.model)
@@ -248,7 +258,7 @@ def classify_articles(
             conn.commit()
         if progress is not None:
             progress(index, len(rows), row["id"])
-    return results
+    return ClassifyReport(results=results, attempted=len(rows), succeeded=len(results), skipped=skipped)
 
 
 def parse_classifier_json(content: str) -> dict: