Support pgvector's `halfvec` dtype

[rsomani95]

I have a table with fp16 embeddings:

\d+ table;


              Column              |           Type           |
----------------------------------+--------------------------+
embedding                         | halfvec(640)             |
...

But when I try to build the index on using CREATE INDEX ON table USING diskann (embedding);, I run into the following error:

ERROR:  data type halfvec has no default operator class for access method "diskann"
HINT:  You must specify an operator class for the index or define a default operator class for the data type.

Are fp16 vectors not yet supported?

[cevian]

Fp16 is not yet supported. We will consider this feature request during our planning, or you are welcome to submit a PR

[npip99]

@cevian Small request that would be nice (If possible), would be to support 4k dimensions if/when FP16 is supported.

OpenAI's text-embedding-3-large would then be able to fit.

[cho-thinkfree-com]

Halfvec is likely to work very well when a database has to store and process a lot of vector data without losing too much precision.

[tjgreen42]

Note that we are already using quantization to compress vectors in the index, so query performance would likely not be boosted much by using f16. (Storage costs in the main table would be reduced, of course).

[xieydd]

VectorChord already support float16 in 0.2.0 @rsomani95 @cho-thinkfree-com.
Furthermore VectorChord enabling support for vectors with over 2000 dimensions, up to 16000 dim. @npip99

[alejandrodnm]

pgvectorscale's diskann release 0.6.0 added supports for up to 16k dimensions #181

[BryceBlankinship]

I have a use case for ~140M 1024-dim vectors that I'd like halfvec for when I inevitably re-embed later on for higher domain-specific benchmark accuracy. That being said, I may attempt this PR in the near future and bring halfvec support to vectorscale.