expose costs in ai chat response

[P3il4]

add usd_cents after a puter.ai.chat call:

// non-streaming
"usage": {
    "prompt_tokens": 12,
    "completion_tokens": 10,
    "cached_tokens": 0,
    "usd_cents": 0.002352
},

// streaming
{"type":"text","text":"hi."}
{"type":"usage","usage":{"prompt_tokens":11,"completion_tokens":9,"cached_tokens":0,"usd_cents":0.000705}}

usage is same as from puter.auth.getMonthlyUsage:

// from the non stream case
"deepseek:deepseek-chat:prompt_tokens": {
    "units": 12,
    "cost": 672,
    "count": 1
},
"deepseek:deepseek-chat:completion_tokens": {
    "units": 10,
    "cost": 1680,
    "count": 1
},
// 672 + 1680 = 2352 / 1_000_000 = 0.002352

ai.prompt.cost-calculated event shares the same logic for consistency
testing:

puter.ai.chat([{
        role: 'user',
        content: `message`,
        cache_control: { type: "ephemeral" } // mind the model minimal cache token length for testing cache cases
    },
], options={model: "claude-haiku-4-5"})

tested on openai (with, without cache), claude (with cache), openrouter (special "billedUsage" case) on live keys, deepseek/xai/gemini/mistral on mock servers
closes #2892

[Salazareo]

can we call them microCents instead of Ucents

I know thats its meant to be the mu micro symbol, but bit less legible than microCents

[P3il4]

can we call them microCents instead of Ucents

I know thats its meant to be the mu micro symbol, but bit less legible than microCents

updated, the rest of the file + event uses ucents though, so this causes a bit of a naming inconsistency

[Salazareo]

can we call them microCents instead of Ucents
I know thats its meant to be the mu micro symbol, but bit less legible than microCents

updated, the rest of the file + event uses ucents though, so this causes a bit of a naming inconsistency

well update as we touch it I guess, I didn't realize rest was also using ucents

[P3il4]

ignore that, it's back at efd7b19