fix: Add List[List[str]] formats for overlapped items in theme extration (Continuation in #2347) (#2355)

## Issue Link / Problem Description
<!-- Link to related issue or describe the problem this PR solves -->
- I updated the _extract_themes_from_overlaps in #2347 a few days ago,
then test on a small documents set and generated examples successfullly.
When I use a knowlege graph created by a large documents set in work for
geneation, it stucks in scenario generation because it feed empty themes
to LLM.
<img width="700" height="500" alt="image"
src="https://github.com/user-attachments/assets/44a13723-3318-48bf-93bb-ff9633e59a6c"
/>

- So a update on Fixes #2347 
- MultiHopSpecificQuerySynthesizer hangs indefinitely during scenario
generation when overlapped items are returned as lists instead of tuples
- The _extract_themes_from_overlaps method only handled tuples and
strings, but overlapped entity pairs can also be returned as lists,
causing no themes to be extracted and the generation process to stall.
- **Overlapped_items is in LIst[List] fromat in my case**. So where is
the code/feature to decide what format will be ? I am not sure. but now
if I take this format into considersion, it can fit in my case.

<img width="1072" height="96" alt="image"
src="https://github.com/user-attachments/assets/95a81826-bd89-4aca-843a-e9fad9fafd44"
/>


## Changes Made
<!-- Describe what you changed and why -->
- Modified _extract_themes_from_overlaps in
src/ragas/testset/synthesizers/multi_hop/specific.py to handle list type
in addition to tuple and str
- Changed condition from isinstance(item, tuple) to isinstance(item,
(tuple, list)) to properly extract entities from list-formatted
overlapped items


## Testing
```
from ragas.llms import LangchainLLMWrapper
from ragas.embeddings import LangchainEmbeddingsWrapper
from ragas.testset import TestsetGenerator
from langchain_community.document_loaders import DirectoryLoader, TextLoader
from langchain_openai import AzureOpenAIEmbeddings, AzureChatOpenAI
from ragas.testset.synthesizers import (
    MultiHopSpecificQuerySynthesizer,
    SingleHopSpecificQuerySynthesizer,
)

import os

# Set your Azure OpenAI credentials
AZURE_ENDPOINT = 
AZURE_API_KEY = 
AZURE_API_VERSION =
AZURE_DEPLOYMENT = 
EMBEDDING_DEPLOYMENT =

# Initialize embeddings
embeddings = AzureOpenAIEmbeddings(
    azure_endpoint=AZURE_ENDPOINT,
    api_key=AZURE_API_KEY,
    api_version=AZURE_API_VERSION,
    azure_deployment=EMBEDDING_DEPLOYMENT
)

# Initialize LLM with JSON mode enabled
llm = AzureChatOpenAI(
    azure_endpoint=AZURE_ENDPOINT,
    api_key=AZURE_API_KEY,
    openai_api_version=AZURE_API_VERSION,
    azure_deployment=AZURE_DEPLOYMENT,
    temperature=0.3,
    timeout=60,
    #model_kwargs={
    #     "response_format": {"type": "json_object"}  # Force clean JSON output
    # }
)

generator_llm = LangchainLLMWrapper(llm)
generator_embeddings = LangchainEmbeddingsWrapper(embeddings)

distribution = [
    # (SingleHopSpecificQuerySynthesizer(llm=generator_llm), 1.0),
    (MultiHopSpecificQuerySynthesizer(llm=generator_llm), 1),
]

# print(distribution)

from ragas.testset.graph import KnowledgeGraph

kg = KnowledgeGraph.load("XXXX.json")

print(f"Knowledge graph loaded with {len(kg.nodes)} entities and {len(kg.relationships)} relations.")

generator = TestsetGenerator(llm=generator_llm, embedding_model=generator_embeddings, knowledge_graph=kg)

async def generate():
    # generate testset
    testset = generator.generate(
        testset_size=3, 
        query_distribution=distribution,
        with_debugging_logs=True,
        raise_exceptions=True   
    )
    print(testset)

    testset.to_pandas().to_csv("test_theme.csv", index=False)

import asyncio
asyncio.run(generate())
```

## References
<!-- Link to related issues, discussions, forums, or external resources
-->
- Related issues: 
- Documentation: 
- External references: 

## Screenshots/Examples (if applicable)
<!-- Add screenshots or code examples showing the change -->

<img width="3127" height="320" alt="image"
src="https://github.com/user-attachments/assets/fc4836d6-4254-4882-b7c7-2c8489ab54bc"
/>

<img width="1080" height="277" alt="image"
src="https://github.com/user-attachments/assets/f2f557bb-90fa-4bfd-b2bf-5354bd4f478d"
/>


---
<!-- 
Thank you for contributing to Ragas! 
Please fill out the sections above as completely as possible.
The more information you provide, the faster your PR can be reviewed and
merged.
-->

Co-authored-by: Kenzo Yan <kenzo.yan@granlund.fi>

Author: kenzoyan

src/ragas/testset/synthesizers/multi_hop/specific.py

@@ -120,6 +120,7 @@ def _extract_themes_from_overlaps(self, overlapped_items: t.Any) -> t.List[str]:
 
         Handles multiple formats:
         - List[Tuple[str, str]]: Entity pairs from overlap detection
+        - List[List[str]]: Entity pairs as lists
         - List[str]: Direct entity names
         - Dict[str, Any]: Keys as entity names
         """
@@ -131,7 +132,7 @@ def _extract_themes_from_overlaps(self, overlapped_items: t.Any) -> t.List[str]:
 
         unique_entities = set()
         for item in overlapped_items:
-            if isinstance(item, tuple):
+            if isinstance(item, (tuple, list)):
                 # Extract both entities from the pair
                 for entity in item:
                     if isinstance(entity, str):