Pull out L2Denorm from TurboQuant

[connortsui20]

Summary

Tracking issue: #7297

This is the first step in decomposing TurboQuant as 2 scalar fns on top of a dictionary.

We would like to (soon) decompose TurboQuant into a Scheme that roughly constructs the following array tree:

L2Denorm(
    norms,
    SorfTransform(
        rotation_signs,
        Dict(
            codes, (indices into centroids)
            values, (centroid book)
        ),
    ),
)

This is the outer layer.

Testing

The logic is the same so existing tests suffice.

[codspeed-hq]

Merging this PR will not alter performance

✅ 1122 untouched benchmarks
⏩ 1530 skipped benchmarks¹

---

_{Comparing ct/tq-norm (9cb5c35) with develop (0e9c0c5)}

Footnotes

1. 1530 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩