Extract URLs from text

Trailing sentence punctuation (period, comma, semicolon, colon, exclamation, question mark) is trimmed off every link. A closing parenthesis, bracket, or brace goes only when the link holds no matching opener, so a Wikipedia-style title like Go_(programming_language) keeps its parenthesis. Deduplication compares full URL strings letter for letter, so case differences and trailing-slash variants all count as separate links. Emails and plain domains without www. are not links and stay out.