We can no longer rely on confidence from chardet since its always 1 with the move to the C based chardet library

So for files where we assume utf-8, use utf-8 if no explicit encoding is
found. Fixes #1993029 [Apostrophe in book title turns into "à€™" upon import](https://bugs.launchpad.net/calibre/+bug/1993029)

src/calibre/ebooks/chardet.py

@@ -154,6 +154,11 @@ def detect_xml_encoding(raw, verbose=False, assume_utf8=False):
             encoding = encoding.decode('ascii', 'replace')
             break
     if encoding is None:
+        if assume_utf8:
+            try:
+                return raw.decode('utf-8'), 'utf-8'
+            except UnicodeDecodeError:
+                pass
         encoding = force_encoding(raw, verbose, assume_utf8=assume_utf8)
     if encoding.lower().strip() == 'macintosh':
         encoding = 'mac-roman'