better error reports for chat fallbacks

[P3il4]

improves error reporting for .chat calls
currently if user sends something invalid like

await puter.ai.chat(
    [],
    { model: "google/gemini-2.5-flash-lite" }
);

we simply respond with

{"success":false,"error":"no fallback model available"}

which isn't really helpful
with this:

{
    "success": false,
    "error": {
        "attempts": [
            {
                "model": "gemini-2.5-flash-lite",
                "provider": "gemini",
                "error": "400 status code (no body)"
            },
            {
                "model": "openrouter:google/gemini-2.5-flash-lite",
                "provider": "openrouter",
                "error": "400 Input required: specify \"prompt\" or \"messages\""
            }
        ],
        "code": "ai_chat_all_providers_failed",
        "$": "heyputer:api/APIError",
        "message": "All AI chat providers failed (2 attempts)",
        "status": 502
    }
}

list all providers tried and what they reported. openrouter makes it clear what is wrong
or if something is wrong on our side, makes it easier to debug from user reports:

...
{
  "model": "openrouter:google/gemini-2.5-flash-lite",
  "provider": "openrouter",
  "error": "403 Key limit exceeded (total limit). Manage it using https://openrouter.ai/settings/keys"
}
...