feat: support vector search/similarity as UDFs

[ChunxuTang]

This PR implements the basic vector search functionalities as datafusion UDFs. It supports queries like:

-- Simple k-NN search
MATCH (p:Person)
WHERE vector_distance(p.embedding, [0.1, 0.2], cosine) < 0.5
RETURN p.name
LIMIT 10

-- Order by similarity
MATCH (p:Person)
RETURN p.name, vector_distance(p.embedding, [0.1, 0.2], l2) AS dist
ORDER BY dist ASC
LIMIT 10

-- Hybrid: Property filter + vector search
MATCH (p:Person)
WHERE p.age > 25
  AND vector_similarity(p.embedding, [0.1, 0.2], cosine) > 0.8
RETURN p.name

See the test_vector_search.rs file for more query examples.

Meanwhile, it adds the parsing of parameters ($query_vector), so later we could also support passing in a parameter for comparison.

Note: Currently, it only works with in-memory data. We'll add the support of Lance ANN indexes as a following-up task.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 88.07339% with 117 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
rust/lance-graph/src/datafusion_planner/udf.rs	85.87%	62 Missing ⚠️
...t/lance-graph/src/datafusion_planner/vector_ops.rs	89.03%	25 Missing ⚠️
rust/lance-graph/src/parser.rs	91.95%	14 Missing ⚠️
rust/lance-graph/src/semantic.rs	90.09%	11 Missing ⚠️
rust/lance-graph/src/simple_executor/expr.rs	0.00%	3 Missing ⚠️
...t/lance-graph/src/datafusion_planner/expression.rs	92.30%	2 Missing ⚠️

📢 Thoughts on this report? Let us know!