Feature Request: Support compound / partitioned VECTOR INDEX (vector + scalar column)

It would be very useful if Azure SQL's `VECTOR INDEX` could support compound (composite) vector indexes, e.g., indexing on (Embedding, PartitionId) so that nearest neighbor search (`VECTOR_SEARCH`) can be constrained by a scalar column.

This would be useful for:
- making multiple separate search spaces that still share the same table/schema.
- avoid post filtering after ANN search

Right now the only workaround is to create separate tables per partitionId, adding schema and app logic complexity.

Proposed Idea:
Allow `CREATE VECTOR INDEX` to accept multiple columns like:
```
CREATE VECTOR INDEX vec_idx ON [dbo].[wikipedia_articles]([title_vector, partition_id])
WITH (METRIC = 'cosine', TYPE = 'diskann');
```

No repository file, test, or entry point is identified. Start by reviewing Azure SQL's CREATE VECTOR INDEX and VECTOR_SEARCH support, then determine whether this samples repository can demonstrate the proposed compound (Embedding, PartitionId) behavior; done would require documented, working support for the proposed syntax.

Written by the indexing model from the issue text.