feat: add full/sliced file reading and large file summarization

[spelech]

Summary

This PR implements full and line-sliced file retrieval (read_file), automated and on-demand large file LLM summarization (summarize_file), and configurable thresholds in the settings UI.

Key Changes

1. FileReaderService & read_file Tool:
   • Securely reads full files or line slices (start_line, end_line, max_lines) across watched paths (indexed_paths) and local storage (LOCAL_STORAGE_PATH).
   • Path traversal prevention (.., null bytes, symlink breakout) and binary file detection.
   • FastMCP tool read_file and REST endpoint GET /admin/api/files/read.
2. SummarizerService & summarize_file Tool:
   • Auto-summarizes files exceeding configurable threshold (default: >500 KB) using LiteLLM chat completions.
   • Caches summaries in SQLite (file_summaries.summary_text) and indexes them as vector documents (doc_type="summary") so large files remain discoverable via hybrid search.
   • FastMCP tool summarize_file and REST endpoint POST /admin/api/files/summarize.
3. Configurable Thresholds in Settings:
   • UI controls for auto-summarization toggle, threshold (KB), max file size ceiling (MB), max lines per read, and chat model.
   • Endpoints: GET/POST /admin/api/settings/files.
   • New FileSettings frontend component and rebuilt production bundle.

Testing & Verification

• 525 backend pytest tests passing (including 29 new tests across settings, reader, summarizer, MCP tools, and indexing integration).
• 291 Vitest frontend tests passing (including new FileSettings.test.tsx).
• Rebuilt frontend production bundle with TypeScript clean pass.