fix(attachments,usage): validate content behind a BOM, judge by the live parser table

Review follow-ups.

A BOM told the sniff which encoding to read, but was also taken as the
verdict: three prepended bytes let any binary through, including as
notes.txt. A BOM now only selects the test — UTF-8 falls through to the
byte rules on the remainder, UTF-16/32 decode and judge the characters
(NUL, unprintable, or replacement chars from bytes the decoder could not
read). Real Notepad-Unicode text still passes, mp4-behind-a-BOM does not,
in either language.

The gate treated the full parser table as a given, but without docling the
fallback extractor has no .tif/.tiff/.bmp/.webp/.vtt/.xml handler, so those
suffixes skipped the content check and reached the plain-text fallthrough —
the original bug, one install away. The worker now passes the keys of the
extractor it actually built, making the second gate stricter than the
route's static one rather than a copy of it.

Cache bins: extraction coerced a missing value to 0 and only non-zero bins
were recorded, so a provider reporting cached_tokens=0 persisted as NULL —
indistinguishable from "not reported", and OpenAI reports exactly that on
every uncached request. Bins are now carried as Optional and recorded when
not None, which is what the nullable columns and the NULL-means-unknown
comment already assumed. Anthropic's cache_read/cache_creation bins had the
same shape and are fixed alongside; the int-or-None coercion is shared in
llm/base.py.

tests/llm/test_openai_responses.py

@@ -1583,5 +1583,26 @@ def test_record_responses_metadata_captures_cache_write_bin(monkeypatch):
         ),
     ))
     # A pure cache-write turn (first request of a prefix) has reads=0 and
-    # writes>0; the details must still be recorded.
-    assert llm._last_usage["prompt_tokens_details"] == {"cache_write_tokens": 9}
+    # writes>0; both bins were reported, so both are recorded — a reported
+    # zero is "no cache hits", not "unknown".
+    assert llm._last_usage["prompt_tokens_details"] == {
+        "cached_tokens": 0,
+        "cache_write_tokens": 9,
+    }
+
+
+@pytest.mark.unit
+def test_record_responses_metadata_omits_unreported_cache_bins(monkeypatch):
+    """A provider that says nothing about caching must persist as NULL, not 0."""
+    llm = _make_llm(monkeypatch, _responses_caps())
+    llm._record_responses_metadata(_ns(
+        id="resp_2",
+        usage=_ns(
+            input_tokens=10,
+            output_tokens=7,
+            total_tokens=17,
+            input_tokens_details=None,
+            output_tokens_details=None,
+        ),
+    ))
+    assert "prompt_tokens_details" not in llm._last_usage