fix: handle case where google model lacks description

[DinoChiesa]

This fixes #325 .

The chatgpt-shell-google--fetch-model-versions fn loads model descriptions from Google via a GET to https://generativelanguage.googleapis.com/v1beta/models

The response is a set of records like the following.

    {
      "name": "models/gemini-2.0-flash-exp",
      "version": "2.0",
      "displayName": "Gemini 2.0 Flash Experimental",
      "description": "Gemini 2.0 Flash Experimental",
      "inputTokenLimit": 1048576,
      "outputTokenLimit": 8192,
      "supportedGenerationMethods": [
        "generateContent",
        "countTokens",
        "bidiGenerateContent"
      ],
      "temperature": 1,
      "topP": 0.95,
      "topK": 40,
      "maxTemperature": 2
    },

In cases in which the model is deprecated, the description field has contained the word "deprecated". so the logic in
chatgpt-shell-google--fetch-model-versions filters out such records. In some cases, the data returned by Google does not include any description. Eg

    {
      "name": "models/gemma-3-27b-it",
      "version": "001",
      "displayName": "Gemma 3 27B",
      "inputTokenLimit": 131072,
      "outputTokenLimit": 8192,
      "supportedGenerationMethods": [
        "generateContent",
        "countTokens"
      ],
      "temperature": 1,
      "topP": 0.95,
      "topK": 64
    },

This change handles that case. With this change, the current Gemma model can be loaded.

I observed a separate problem using the Gemma model. It returns

Developer instruction is not enabled for models/gemma-3-27b-it

This is a matter of API usage. See #334 .

[xenodium]

Thank you for the fix!