fix: correctly merge update_community results from semaphore_gather

[sprotasovitsky]

Summary

• Fixes tuple unpacking crash in add_episode() when update_communities=True
• semaphore_gather returns a list of N tuples (one per node), not a single tuple — the results must be merged via iteration

Problem

update_community() returns tuple[list[CommunityNode], list[CommunityEdge]] per node. semaphore_gather() collects these into a list:

# semaphore_gather returns: [(nodes_A, edges_A), (nodes_B, edges_B), (nodes_C, edges_C)]
# Old code tried:
communities, community_edges = await semaphore_gather(...)
# This only works when exactly 2 nodes are in the episode — crashes otherwise

This bug was introduced in v0.19.0 (commit dcc9da3) when community results were added to AddEpisodeResults. Every version since v0.19.0 crashes with ValueError: too many values to unpack (expected 2) when update_communities=True and the episode has != 2 nodes.

Fix

community_results = await semaphore_gather(...)
for comm_nodes, comm_edges in community_results:
    communities.extend(comm_nodes)
    community_edges.extend(comm_edges)

Test plan

• Verified with episodes producing 3-7 nodes — no crash
• Communities are correctly updated and persisted to Neo4j
• AddEpisodeResults correctly contains merged community data

Fixes #836

[danielchalef]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[sprotasovitsky]

I have read the CLA Document and I hereby sign the CLA

[danielchalef]

Thanks for this — same root cause as #1486 (update_community results returned by semaphore_gather aren't merged/flattened in add_episode). Consolidating on #1486 (includes a regression test). Closing as superseded by #1486.