Integrate GPT OSS Safeguard into ThreatExchange as chat command

[ZhiyLiu]

No description provided.

[Dcallies]

Let's chat offline, we should be using the classifier interface for this instead.

[ZhiyLiu]

Test Plan:
export openai key

threatexchange classify modapi -s 'I hate you'
Expected result:
{
"parsed": {
"action": "allow",
"category": "none",
"confidence": 0.97,
"label": "ALLOW",
"rationale": "The statement is a simple expression of dislike without targeting protected groups or containing disallowed content."
},
"raw_text": "{"label":"ALLOW","action":"allow","category":"none","confidence":0.97,"rationale":"The statement is a simple expression of dislike without targeting protected groups or containing disallowed content."}"
}

[Dcallies]

Hey @ZhiyLiu , I'm a bit backlogged, but I will come back to this when I have free time to try and clean it up for merge!