Enhanced quantization (oQe)

This might be a really silly question, however I notice that you haven't included any of the Enhanced quantization (oQe) levels.
My understanding from [https://github.com/jundot/omlx/blob/main/docs/oQ_Quantization.md](https://github.com/jundot/omlx/blob/main/docs/oQ_Quantization.md) is that these enhanced quants add more precision, especially to the lower quants.
Do you plan to include these in the future, or have I misunderstood what oQe is?

Start by reading the linked docs/oQ_Quantization.md to understand the oQe levels and how they differ from the quantizations currently evaluated. Then inspect mlx-eval's quantization-related entry points and document the required scope; this issue is done only when the supported levels and evaluation expectations are agreed.

Written by the indexing model from the issue text.