Fix: clean_value returns "|foo" from link with 3+ parameters

Issue found when parsing "[[Fichier:...|...|...]]" image links
from French WikiPEDIA. Because 'Fichier' was not recognised
as a skipped prefix (needs to be addressed), the link was handled
by the default regex handler, which never used the value of
`m.group(4)`, because only File: and Image: 'links' actually
have that third parameter; this is why this was not caught.

XXX: Localized data to replace "File:" and "Image:" prefixed
with "Fichier:" etc.
XXX: A toggle in clean_value and clean_node whether to collect
image link data?

src/wiktextract/clean.py

@@ -1359,7 +1359,7 @@ def repl_link_bars(m: re.Match) -> str:
         lnk = m.group(1)
         if re.match(r"(?si)(File|Image)\s*:", lnk):
             return ""
-        return clean_value(wxr, m.group(4) or m.group(2) or "", no_strip=True)
+        return clean_value(wxr, m.group(5) or m.group(2) or "", no_strip=True)
 
     def repl_1_sup(m: re.Match) -> str:
         return to_superscript(clean_value(wxr, m.group(1)))