fix: handle empty SPOKE API response in get_context_using_spoke_api

[octo-patch]

Fixes #42

Problem

When the SPOKE Knowledge Graph API returns an empty list for a node (e.g. neurofibromatosis 2 which may have been removed in a KG update), get_context_using_spoke_api raises an IndexError at line 128:

context += node_value + " has a " + node_context[0]["data"]["properties"]["source"] + ...

This crashes because node_context[0] is accessed without checking whether node_context is empty.

Solution

Wrap the identifier-appending line in a guard that only executes when node_context is non-empty:

if node_context:
    context += node_value + " has a " + node_context[0]["data"]["properties"]["source"] + ...

When the API returns no results for a node, the function now gracefully returns an empty context string and an empty DataFrame instead of crashing.

Testing

Added unit tests in kg_rag/test/test_spoke_api_empty_response.py that mock the SPOKE API:

1. test_empty_node_context_does_not_raise — verifies that an empty API response no longer raises IndexError, and that the function returns an empty context and empty DataFrame.
2. test_non_empty_node_context_includes_identifier — verifies that normal (non-empty) API responses still include the identifier information in the returned context.

Both tests pass with the fix applied.