feat(retrieval): add file-level chunked vectorization

[mildred522]

Description

This PR contributes to Category 3 ("Embedding Processing & Chunking Strategy") in the token/cost optimization tracker
described in #744.

It focuses on the file-level part of that work. Previously, long text files could be indexed as a single
coarse file-level embedding, which reduced retrieval quality for oversized documents and left file-level
chunking behavior undefined.

This change adds configurable chunked vectorization for long text files and collapses chunk-level hits back
to the base file URI during retrieval, so long files gain finer-grained vector coverage internally while
retrieval results remain file-level externally.

Related Issue

• Related to [Tracking]: Token 消耗与成本优化问题汇总 / Token Consumption & Cost Optimization Tracker #744
• Related to Embedding truncation and chunking should have clearer responsibilities across memory, file, and directory vectorization #531

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Refactoring (no functional changes)
• Performance improvement
• Test update

Changes Made

• Add file_chunk_chars and file_chunk_overlap config options, plus validation in OpenVikingConfig
• Add chunked vectorization for long text files in embedding_utils.py
• Collapse chunk-level retrieval hits back to base file URIs in HierarchicalRetriever
• Update English and Chinese configuration docs for the new file chunking settings
• Add regression tests for config validation, long-file chunked vectorization, and retrieval-side chunk collapse

Testing

• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• I have tested this on the following platforms:
  • Linux
  • macOS
  • Windows

Targeted local verification completed:

• py -3.11 -m ruff check openviking_cli\utils\config\open_viking_config.py openviking\utils\embedding_utils.py openviking\retrieve\hierarchical_retriever.py tests\misc\test_openviking_config_file_chunking.py tests\misc\test_file_chunk_vectorization.py tests\retrieve\test_hierarchical_retriever_chunk_collapse.py
• py -3.11 -m pytest tests\misc\test_openviking_config_file_chunking.py tests\misc\test_file_chunk_vectorization.py tests\retrieve\test_hierarchical_retriever_chunk_collapse.py -q

Checklist

• My code follows the project's coding style
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• Any dependent changes have been merged and published

Additional Notes

This PR is intentionally scoped to file-level chunked vectorization only.
It does not attempt to define chunking behavior for memory or directory indexing.

Chunk-level candidates are currently collapsed back to file-level results using the generated
chunk URI suffix convention, while preserving source_uri in metadata.

[github-actions]

Failed to generate code suggestions for PR

[mildred522]

This PR overlaps with #858 in problem space, but the approach is different.

#858 focuses on configurable truncation / text source selection to avoid oversized embedding inputs.
This PR focuses on file-level chunked vectorization, so long text files keep finer-grained vector coverage, while
retrieval still returns file-level results by collapsing chunk hits back to the base URI.

If maintainers prefer the smaller config-only step first, I can rebase this PR later.