Perplexity colors extension updates

[SeanScripts]

Checklist:

• I have read the Contributing guidelines.

Just some bugfixes for perplexity colors for better stability. It seems to be working well now, but let me know if you find any bugs if you test it out. I think it's very interesting to see what happens when you choose to sample a different token at some point in a message.

• The "continue" functionality works better with this now, and the last sampled token is highlighted if it can be found at the end of the message, when it wasn't colored at all before.
• Clicking a token option to continue generation from there works more consistently.
• Resolved a bug where sometimes parts of past messages would get removed because of non-unique IDs.
• Added a check for missing tokens that seems to mostly resolve the issue where you start a new chat and it tries to highlight tokens from the greeting message that were in the last message of the previous chat, though ideally we could clear the logits processor somehow when starting a new chat.

Remaining issues:

• Sometimes particular tokens still can't be found in the message, so they end up not getting highlighted. This tends to happen with certain special characters in different writing systems.
• Display of <tool_call>...</tool_call> gets messed up after generation and replaced with whitespace when adding the colors. Might be related to the HTML escape step that is done. This one is my next priority to figure out. Perplexity measurements with tool calls are a great way to see how confident the model is in using a particular tool and what arguments it's using.
• Continuing generation after clicking a token with newlines or other whitespace characters often won't work as intended, because the whitespace will get trimmed before generation.