Ouro llm support for candle

```
Exceptional Parameter Efficiency

Through 7.7T token pre-training, our models demonstrate remarkable parameter efficiency:

Ouro-1.4B: Matches the performance of 4B standard transformer models
Ouro-2.6B: Rivals 8B standard transformer models
Achieves 2-3× parameter efficiency improvements across diverse benchmarks
```
Reference: https://ouro-llm.github.io/

Start by reading the Ouro reference at https://ouro-llm.github.io/ and reviewing Candle's existing model-support entry points. Clarify the required Ouro variants, interfaces, weights, and validation expectations before implementation; done means the requested Ouro models are supported in Candle with demonstrated loading and inference.

Written by the indexing model from the issue text.