fix: filter None chunk results in graph merge_batch_data

[dex0shubham]

Problem

In one-stage extraction, the single-chunk path does:

graph = self.data_extractor.invoke(inp)
graph_list = [graph]
... merge_batch_data(graph_list)

with_structured_output(method="function_calling") can return None when the model emits no tool call. merge_batch_data([None]) then falls through to the tuple branch and hits assert len(data_list_or_tuple) == 2, raising AssertionError instead of degrading gracefully. (The multi-chunk batch path already filters None via _filter_none_results; the single-chunk path did not.)

This affects AutoGraph, AutoHypergraph, and the three specialized graphs (AutoTemporalGraph, AutoSpatialGraph, AutoSpatioTemporalGraph), which all inherit merge_batch_data.

Fix

Filter None entries from the list form at the top of merge_batch_data (in graph.py and hypergraph.py — the only two definitions). An all-None list returns an empty graph. The tuple form ((nodes_lists, edges_lists)) is a tuple, not a list, so it's unaffected. Fixing this one chokepoint covers all five graph types.

Tests

merge_batch_data([None]) returns an empty graph/hypergraph instead of raising, for both AutoGraph and AutoHypergraph. Both fail with AssertionError on the pre-fix code.

ruff check/ruff format --check on hyperextract clean.