Abnormal-AT-segment-identification fixing

[WhiteFeather127]

fix #7

[TaranDahl]

_find_full_name_mentions 对每个 text segment 都会遍历全部缓存成员做 re.finditer。如果群组成员达到上千人，每条含 @ 的文本都会产生 O(n) 的正则扫描。对于高频的 QQ/Discord 消息转发场景可能有性能影响。
_find_full_name_mentions iterates through all cached members and runs re.finditer for every text segment. If a group has thousands of members, every message containing an @ triggers an O(n) regular expression scan. This may cause performance issues in high-frequency QQ/Discord message forwarding scenarios.

[WhiteFeather127]

已修复，详见 commit e5386ec。

优化了 _find_full_name_mentions 的性能问题：

• 将逐成员 O(n) 正则扫描改为单个组合交替正则 O(1)
• 使用 casefold() 做大小写不敏感的 display name 查找
• 长名字优先匹配，避免短前缀误抢

---

Fixed in e5386ec.

Optimized _find_full_name_mentions performance:

• Changed from per-member O(n) regex scans to a single combined alternation regex (O(1))
• Uses casefold() for case-insensitive display name lookup
• Longer names match first to prevent short-prefix false matches