fix: strip server-assigned ids before re-uploading annotations

`job.annotations()` returns existing annotations with their server-assigned
`id` fields. When those are re-submitted via `set_annotations` (a full-replace
PUT), CVAT rejects/mishandles them because it treats `id` as the identity of
an existing object. Strip `id` from tags, shapes, and tracks in
`JobAnnotations.request()` so the server creates them fresh.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Author: FelixAbrahamsson

next_cvat/client/job_annotations.py

@@ -84,10 +84,22 @@ def add_tag_(
         return self
 
     def request(self) -> models.LabeledDataRequest:
+        def remove_id_from_dict(item_dict):
+            """Remove 'id' field from dictionary if present"""
+            if isinstance(item_dict, dict):
+                return {k: v for k, v in item_dict.items() if k != "id"}
+            return item_dict
+
         request = models.LabeledDataRequest()
         request.version = self.annotations["version"]
-        request.tags = self.annotations["tags"]
-        request.shapes = self.annotations["shapes"]
-        request.tracks = self.annotations["tracks"]
+
+        # Remove ID fields from existing annotations
+        request.tags = [remove_id_from_dict(tag) for tag in self.annotations["tags"]]
+        request.shapes = [
+            remove_id_from_dict(shape) for shape in self.annotations["shapes"]
+        ]
+        request.tracks = [
+            remove_id_from_dict(track) for track in self.annotations["tracks"]
+        ]
 
         return request