fix: correct impersonation detection logic and improve tweet container selector reliability

[brenopolanski]

Summary

Note: Merge this PR first #63

This PR fixes a critical bug in the impersonation detection logic that was incorrectly flagging legitimate tweets as impersonation attempts. Additionally, it improves the robustness of tweet container selectors to handle different Twitter/X localizations and UI variations.

Changes

• Fixed incorrect flagging: The previous logic was comparing the impersonator's handle with the safe handle and then flagging tweets matching the safe handle (the legitimate account), which was backwards
• Correct behavior: Now properly identifies and flags tweets from the impersonator's handle instead of the legitimate user's tweets
• Added fallback selectors for localized versions of Twitter/X (e.g., Portuguese "Conversa", Japanese "会話")
• Implemented graceful degradation: if timeline container isn't found, searches entire document for tweets

Preview