fix: close adversarial review findings on document parsing

Malformed download URLs are rejected before the second request, error bodies are read only up to a bound, the telemetry summary keeps structural dispatch for duck-typed results, and the BMP magic now requires a valid DIB header. Docs and design are aligned with the implementation, and new test wording is neutral.

tests/unit/test_ocr_client.py

@@ -5,6 +5,7 @@ retry_exhausted/circuit_open/stalled 三组断言即设计 §6 ③ 的 G1 契约
 """
 
 import asyncio
+from types import SimpleNamespace
 
 import httpx
 import pytest
@@ -826,6 +827,21 @@ class TestDocumentTelemetry:
         assert OcrClient._summarize(_LAYOUT_OK) == "<elements n=1 pages=1>"
         assert OcrClient._summarize(_TEXT_OK) == "LINE-1"
 
+    def test_summary_keeps_structural_dispatch_for_duck_typed_results(self):
+        """旧实现按结构分派(`hasattr(result, "text")`)。下游测试替身可能返回非库类型的鸭子对象,
+        改成 isinstance 会让它们在开启遥测后以原生 AttributeError 漏出——无登记的回归。"""
+        assert OcrClient._summarize(SimpleNamespace(text="hello", raw={})) == "hello"
+        assert (
+            OcrClient._summarize(SimpleNamespace(elements=[1, 2], page_sizes=[(1.0, 1.0)]))
+            == "<elements n=2 pages=1>"
+        )
+        assert (
+            OcrClient._summarize(
+                SimpleNamespace(elements=[1], page_sizes=[(1.0, 1.0)], markdown="abc")
+            )
+            == "<document elements=1 pages=1 md_chars=3>"
+        )
+
     async def test_failed_attempt_row_carries_the_class_prefix(self):
         recorder = _MemoryRecorder()
         client, _, _ = _client([_src()], [ResultInvalidError("bad package")], telemetry=recorder)