Add metadata-based reranking to search results

[zohaib-7035]

Summary:
This PR introduces a metadata-based reranking feature to improve the relevance of search results in the Knowledge Space Agent. Search results are now scored and reordered based on:

• Publication year: Newer datasets get a small boost.
• Citation count: Highly cited datasets are preferred.
• Trusted sources: Datasets from trusted sources (e.g., Allen Brain Atlas, GENSAT, EBRAINS) get a higher ranking.

Changes:

• Added rerank_results_using_metadata() function in ks_search_tool.py.
• Integrated reranking into general_search and smart_knowledge_search.
• Added debug prints for scores to verify reranking during testing.

Testing:
Tested with sample queries like mouse brain, memory, and hippocampus. Results were correctly reranked according to metadata.

Screenshot of test output:

Impact:

• Improves search relevance for end-users.
• Sets foundation for further enhancements like advanced scoring and weighting.

[QuantumByte-01]

Closing — the reranking score math is incorrect. Adding year/10000 (~0.2) and citations/100 as flat boosts to normalized relevance scores completely distorts the ranking regardless of semantic match quality. A dataset from 2024 with 1000 citations would get +0.2024 + 10.0 boost, drowning out any relevance signal. The feature idea is valid but the scoring needs a proper normalization strategy (e.g. min-max scaling, log normalization). Please rethink the math and reopen.