feat: add support in vector_store for server-side embeddings

[jordanrfrazier]

Adds the $vectorize functionality to the vector store. Still uses the legacy astrapy paths.

Note: does not add functionality to the cache.

[nicoloboschi]

Great work!!
I've left some comments

[hemidactylus]

Very nice and comprehensive work!

My only suggestion concerns the fixtures around testing. With the vectorize feature eventually landing in prod, it could be better to prepare for that (something like a twin set of fixtures, vector- and vectorize-based) and add a small amount of skipifs to momentarily block the latter in prod? For ways to combine this with the pytest parametrize, I found a nice way to condition items of a parametrize (see here).

[jordanrfrazier]

My only suggestion concerns the fixtures around testing. With the vectorize feature eventually landing in prod, it could be better to prepare for that (something like a twin set of fixtures, vector- and vectorize-based) and add a small amount of skipifs to momentarily block the latter in prod? For ways to combine this with the pytest parametrize, I found a nice way to condition items of a parametrize (see here).

Cool, this is a neat technique to know. I did look into using this pattern, but since we have several tests that use one or the other of the StoreEmbedding vs. ParseEmbedding, it was easier to just put a single skipif on the vectorize_store, which can be removed once vectorize is in prod.

We could probably refactor the tests into separate modules for the different embeddings, then use the linked pattern in a further revision.

approved