Add CAGRA-Q build (compression)

[achirkin]

Add a cagra::compress function that implements CAGRA-Q (VQ + PQ) compression of a given dataset.
The result, compressed_dataset, is supposed to complement the CAGRA graph during cagra::search in place of a raw dataset.

Current state:

• The code runs and produces a meaningful output (tested internally by running the original prototype search with the generated compressed dataset); the recall levels are approximately the same as with the prototype implementation.
• No test coverage yet (need to coordinate with the search PR CAGRA-Q search #2206)
• Full pq_bits support ([4,5,6,7,8] - same as in IVF-PQ)
• Any pq_dim values are accepted, but the dataset is not padded and thus dim must be a multiple of pq_dim.
• The codebook math type is hardcoded to half to match the prototype implementation for now. This could be a runtime (build) parameter as well.
• All common input data types should work (uint8_t, int8_t, half, and float compile), but I tested only float.

[tfeher]

Thanks @achirkin for this PR! Here is a first batch of my comments.

[tfeher]

Many thanks Artem for updating the PR, and also adding serialization methods. Overall it looks good, but I still have an issue with the index types.

[tfeher]

Thanks Artem for integrating data compression into RAFT CAGRA! The PR looks good to me, just have two minor questions below.

Ideally we should go ahead and merge this, so that the follow up PR (#2206) is easier to review.

[tfeher]

Thanks Artem, the PR looks good to me!

[achirkin]

Latest update:

1. Renamed dataset_view() back to dataset() thus making the PR non-breaking.
2. Made a few other smaller renamings as per our discussion
3. Marked dataset() deprecated and added an EXPERIMENTAL note on the new index parameter.

[achirkin]

/merge