Try several encoding.

Even chardet may return a invalid encoding.

By looping on all potential encodings detected by chardet, we are more
tolerant.

src/warc2zim/utils.py

@@ -77,13 +77,19 @@ def to_string(input_: str | bytes, encoding: str | None) -> str:
     # First try declared encoding in content.
     content_start = input_[:1024].decode("ascii", errors="replace")
     if m := ENCODING_RE.search(content_start):
-        encoding = m.group("encoding")
+        encodings = [m.group("encoding")]
     else:
-        encoding = chardet.detect(input_)["encoding"]
+        all_encodings = chardet.detect_all(input_)
+        encodings = [e["encoding"] for e in all_encodings]
 
-    if not encoding:
+    if not encodings:
         raise ValueError(f"Impossible to detect encoding of content {input_[:200]}")
-    return input_.decode(encoding)
+    for encoding in encodings:
+        try:
+            return input_.decode(encoding)
+        except ValueError:
+            pass
+    raise ValueError(f"Impossible to decode content {input_[:200]}")
 
 
 def get_record_content(record: ArcWarcRecord):