add false-positive filter and bulk-clear for findings

Active Feedback Rules gets a status filter, and Field Baseline
Deviations gets multi-select with a bulk apply bar. Bulk scope is
restricted to exact-value/field only, so a batch action can never
create a broad rule that would suppress a genuinely new deviation.

src/fgai/feedback.py

@@ -105,6 +105,28 @@ class FeedbackStore:
         self.path.write_text(json.dumps(entries, indent=2), encoding="utf-8")
         return entry
 
+    def add_many(self, items: list[dict[str, object]]) -> dict[str, object]:
+        """Normalize and save a batch of entries with a single read-modify-write.
+
+        One bad item never blocks the rest — each is validated independently and
+        failures are reported by index instead of aborting the whole batch.
+        """
+        added_by_id: dict[str, dict[str, object]] = {}
+        errors: list[dict[str, object]] = []
+        for index, item in enumerate(items):
+            try:
+                entry = _normalize(item)
+                added_by_id[entry["id"]] = entry  # later duplicates in the same batch win, like sequential add() calls would
+            except ValueError as exc:
+                errors.append({"index": index, "error": str(exc)})
+        added = list(added_by_id.values())
+        entries = self.entries()
+        entries = [value for value in entries if value.get("id") not in added_by_id]
+        entries.extend(added)
+        self.path.parent.mkdir(parents=True, exist_ok=True)
+        self.path.write_text(json.dumps(entries, indent=2), encoding="utf-8")
+        return {"added": added, "errors": errors, "remaining": len(entries)}
+
     def revoke(self, entry_id: str) -> dict[str, object]:
         entries = self.entries()
         before = len(entries)