* Add palettized model definitions.

* Adapt to new API

TODO: update SPM package dependency.

* Allow download of private repos with HF token

This is not exposed in the UI.

* Use split_einsum_v2 when available

* Minor README changes.

* iOS: reduceMemory if models not quantized or < 6GB RAM

* Only show quantized models in iOS 17, macOS 14.

* Update ml-stable-diffusion package reference.

* Update URLs of quantized models.