[feat] Automatically add model slugs from requests

[NextDev65]

The model switcher should automatically populate its list of available models by intercepting network requests to discover new model slugs. This will ensure the list is always up-to-date without manual updates to the script.

• inspect network req/resp for model slugs

• implement extraction to list + handle redundant extraction

• find best source of labels, possibly external

• implement extraction to list + handle redundant extraction

[NextDev65]

Now that labels are added (ie: "GPT-5.4 Thinking Mini" for gpt-5-4-t-mini), extract these as well.

• likely from frontend, as it is not apparent in network requests/responses