feat: manage property graphs defined over tables

Replaces #4322, keeping only its property graph defined over tables.

A property graph maps node labels and relationship types to existing tables of its namespace, in the shape of SQL/PGQ's `CREATE PROPERTY GRAPH`. It stores only that definition: a query reads the tables as nodes and edges through SQL `GRAPH_TABLE`, as they are when it runs.

- **Rust:** `Connection::create_property_graph` takes a `PropertyGraphDefinition` of `NodeTable`s and `EdgeTable`s. `describe_property_graph`, `drop_property_graph` and `list_property_graphs` manage it.
- **Python:** `lancedb.graph` has `NodeTable`, `EdgeTable` and `Endpoint`. Sync and async connections gain `create_property_graph`, `open_property_graph`, `drop_property_graph` and `list_property_graphs`, and a `PropertyGraph` handle describes its graph.
- **Support:** remote connections only. Local connections raise `NotImplementedError` in Python and return `NotSupported` in Rust.

```python
import lancedb
from lancedb.graph import EdgeTable, Endpoint, NodeTable

db = lancedb.connect("db://demo", api_key="...")
db.create_property_graph(
    "social",
    nodes=[NodeTable("person", key="person_id", label="Person")],
    edges=[
        EdgeTable(
            "knows",
            label="KNOWS",
            source=Endpoint("src_id", references=("person", "person_id")),
            destination=Endpoint("dst_id", references=("person", "person_id")),
        )
    ],
)
db.execute_query("""
    SELECT * FROM GRAPH_TABLE (social
        MATCH (p:Person)-[:KNOWS]->(f:Person) WHERE p.age > 30
        COLUMNS (p.name AS person, f.name AS friend))
""").read_all()
```

docs/src/python/python.md

@@ -493,6 +493,31 @@ automatic scheduling requests.
 
 ::: lancedb.view.ViewDescription
 
+## Property Graphs
+
+A property graph is defined over tables of its namespace: which tables hold
+nodes and which hold edges, each node table's key and label, and the node keys
+each edge table's source and destination columns reference -- the shape of
+SQL/PGQ's `CREATE PROPERTY GRAPH`. It stores no rows of its own; a query reads
+its tables as nodes and edges through SQL `GRAPH_TABLE`.
+
+Create one with
+[create_property_graph][lancedb.db.DBConnection.create_property_graph], and
+manage it with
+[open_property_graph][lancedb.db.DBConnection.open_property_graph],
+[list_property_graphs][lancedb.db.DBConnection.list_property_graphs] and
+[drop_property_graph][lancedb.db.DBConnection.drop_property_graph].
+
+::: lancedb.graph.NodeTable
+
+::: lancedb.graph.EdgeTable
+
+::: lancedb.graph.Endpoint
+
+::: lancedb.graph.PropertyGraph
+
+::: lancedb.graph.PropertyGraphDescription
+
 ## Expressions
 
 Type-safe expression builder for filters and projections. Use these instead
@@ -840,6 +865,14 @@ Table hold your actual data as a collection of records / rows.
 
 ::: lancedb.materialized_view.AsyncMaterializedView
 
+## Property Graphs (Asynchronous)
+
+`AsyncConnection` returns this handle from
+[create_property_graph][lancedb.db.AsyncConnection.create_property_graph] and
+[open_property_graph][lancedb.db.AsyncConnection.open_property_graph].
+
+::: lancedb.graph.AsyncPropertyGraph
+
 ## Indices (Asynchronous)
 
 Indices can be created on a table to speed up queries. This section