TurboQuant encoding for Vectors

[connortsui20]

Continuation of #7167, authored by @lwwmanning

Summary

Lossy quantization for vector data (e.g., embeddings) based on TurboQuant (https://arxiv.org/abs/2504.19874). Supports both MSE-optimal and inner-product-optimal (Prod with QJL correction) variants at 1-8 bits per coordinate.

Key components:

• Single TurboQuant array encoding with optional QJL correction fields, storing quantized codes, norms, centroids, and rotation signs as children.
• Structured Random Hadamard Transform (SRHT) for O(d log d) rotation, fully self-contained with no external linear algebra library.
• Max-Lloyd centroid computation on Beta(d/2, d/2) distribution.
• Approximate cosine similarity and dot product compute directly on quantized arrays without full decompression.
• Pluggable TurboQuantScheme for BtrBlocks, exposed via WriteStrategyBuilder::with_vector_quantization().
• Benchmarks covering common embedding dimensions (128, 768, 1024, 1536).

Also refactors CompressingStrategy to a single constructor, and adds vortex_tensor::initialize() for session registration of tensor types, encodings, and scalar functions.

API Changes

Adds a new TurboQuant encoding + some other things. TODO

Testing

TODO

[codspeed-hq]

Merging this PR will degrade performance by 29.05%

❌ 2 regressed benchmarks
✅ 1120 untouched benchmarks
⏩ 1530 skipped benchmarks¹

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	take_map[(0.1, 1.0)]	1.6 ms	2.3 ms	-29.05%
❌	Simulation	take_map[(0.1, 0.5)]	977.4 µs	1,209.5 µs	-19.2%

---

_{Comparing ct/turboquant (77a4288) with develop (e3c7401)}

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. ↩

[connortsui20]

I would say this is ready for review now, but only with respect to the structure. I have yet to go through the implementation and make sure things make sense, but I have made it so the structure makes sense and we correctly handle the different floating point type inputs as well as null vectors.

I think it would be good to get a review now, and maybe we should just merge this and iterate later.

[gatesn]

Let's do it

[connortsui20]

I'll rebase and fix the errors tomorrow