Cypher undirected MATCH can bind the seed node as the peer alias on incoming edges

Summary

Direct Cypher execution can bind the seed node as the peer alias for an undirected relationship pattern when the stored edge is incoming.

This showed up while rerunning the LDBC SNB Interactive interactive-short-3 shape on current master / v0.53.10 (7095102df492feed0ce7a83240ee406c779c5f81). The query no longer hard-fails, but it returns semantically wrong rows.

Minimal repro

import pandas as pd
import graphistry

nodes = pd.DataFrame([
    {"id": 1, "label__Person": True, "firstName": "Alice", "lastName": "Seed"},
    {"id": 2, "label__Person": True, "firstName": "Bob", "lastName": "Friend"},
])

edges = pd.DataFrame([
    {"s": 2, "d": 1, "type": "KNOWS", "creationDate": 10},
])

g = graphistry.nodes(nodes, "id").edges(edges, "s", "d")

q = """
MATCH (n:Person {id: $personId })-[r:KNOWS]-(friend)
RETURN
    friend.id AS personId,
    friend.firstName AS firstName,
    friend.lastName AS lastName,
    r.creationDate AS friendshipCreationDate
ORDER BY friendshipCreationDate DESC, toInteger(personId) ASC
"""

res = g.gfql(q, params={"personId": 1}, engine="pandas")
print(res._nodes[["personId", "firstName", "lastName", "friendshipCreationDate"]].to_dict(orient="records"))

Actual

[{"personId": 1, "firstName": "Alice", "lastName": "Seed", "friendshipCreationDate": 10}]

Expected

[{"personId": 2, "firstName": "Bob", "lastName": "Friend", "friendshipCreationDate": 10}]

Benchmark impact

This blocks direct-Cypher correctness for LDBC SNB Interactive interactive-short-3 / IS3 on pyg-bench. GFQL still has a workaround path, but the direct Cypher path is now a semantic mismatch rather than an unsupported query.