performance: optimize PUNCT_TBL creation (#791)

Author: deedy5

trafilatura/deduplication.py

@@ -22,7 +22,7 @@
 
 BIN_COUNT_FUNC = getattr(int, "bit_count", lambda x: bin(x).count("1"))
 
-PUNCT_TBL = str.maketrans({i: ' ' for i in range(0x10FFFF) if unicodedata.category(chr(i)).startswith('P')})
+PUNCT_TBL = str.maketrans({i: " " for i in range(0x10FFFF) if unicodedata.category(chr(i))[0] == "P"})
 
 
 @lru_cache(maxsize=1024)