Implement Indexing for RabitQ

[tvhong]

Notes

Implement the Indexing Phase for RabitQ.

Paper: https://arxiv.org/abs/2405.12497
My summary of the indexing phase: #32 (comment)

Please note that the current implementation is very inefficient since I don't have much experience writing high performance code, but it should at least be correct.

[tyb0807]

Thanks! Just one high level question, why Array1 and Array2 instead of Vec and Vec<Vec>? Asking just for the sake of consistency with other parts of the code base.

[tvhong]

@tyb0807 The querying phase needs to do some matrix multiplication and vector dot products (see #32 (comment)). So, I store these as ndarray objects.

That being said, I'm not set on the data type at the moment. Maybe ndarray doesn't support serialization well, or that we should use nalgebra since we're only dealing with low-dimensionality data. But, regardless, it's quite easy to change the data type once we decide that ndarray isn't the right type.

[hicder]

you'll need to use the normalized vector

[tvhong]

Updated the code to use normalized vectors instead.

[hicder]

LGTM!