feat(models): add lucy-2.5 model (realtime + batch)

[AdirAmsalem]

Makes Decart's latest Lucy generation, lucy-2.5, available to Python SDK users for both realtime streaming and batch/queue video editing. This brings the Python SDK to parity with the other SDKs and lets users adopt the newest model without waiting on a client update.

lucy-2.5 is a canonical model (no deprecation aliases) supporting video editing with an optional reference image — the same input shape as lucy-2.1.

Usage:

from decart import DecartClient, models

client = DecartClient(api_key="...")

# Batch / queue
result = await client.queue.submit_and_poll({
    "model": models.video("lucy-2.5"),
    "prompt": "Watercolor painting style with soft brushstrokes",
    "data": video_file,
})

# Realtime
connection = await client.realtime.connect(
    model=models.realtime("lucy-2.5"),
    ...
)

---

Note

Low Risk
Additive model registry and test coverage only; no changes to auth, request handling, or existing model behavior.

Overview
Adds lucy-2.5 as a canonical model for both realtime streaming and queue/batch video editing, matching the pattern used for lucy-2.1.

Realtime uses /v1/stream at 20 fps and 1088×624 (vs 30 fps for lucy-2.1). Batch jobs use /v1/jobs/lucy-2.5 with the same VideoEdit2Input shape (prompt + optional reference image). The playground CLI model list and model registry tests are updated accordingly.

^{Reviewed by Cursor Bugbot for commit e6618ce. Bugbot is set up for automated code reviews on this repo. Configure here.}