feat: add optional --db flag to persist search results to SQLite

[larock22]

Summary

Add an optional --db <path> CLI flag that persists all search results (and scraped page data when --scrape is used) into a lightweight SQLite database. This enables users to build up a queryable local archive of searches without external dependencies.

Current Behavior

sophon-cli prints search results to stdout and discards them. There is no built-in way to save result metadata, URLs, snippets, or scraped page content for later analysis.

Expected Behavior

When --db results.db is passed:

• Search results are written to results.db immediately after the API response is received.
• If --scrape is also enabled, scraped page content and crawl telemetry are appended to the same database.
• The schema is lightweight and self-contained (SQLite, no external server).

Technical Context

• Entrypoint: src/cli/runner.rs (run_single_provider and run_all_enabled)
• CLI args: src/cli/args.rs
• Result types: src/domain/result.rs (SearchResponse, SearchBatchResponse, SearchResult variants)
• Scrape types: src/cli/scrape.rs (ScrapedSite, ScrapedPage)
• Architecture boundary: DB code belongs in the CLI layer only; domain must remain pure.

Proposed Schema

CREATE TABLE search_runs (
  id INTEGER PRIMARY KEY AUTOINCREMENT,
  query TEXT NOT NULL,
  provider TEXT NOT NULL,
  run_at TEXT NOT NULL
);

CREATE TABLE search_results (
  id INTEGER PRIMARY KEY AUTOINCREMENT,
  run_id INTEGER NOT NULL,
  result_type TEXT NOT NULL,
  title TEXT NOT NULL,
  url TEXT NOT NULL,
  snippet TEXT,
  source TEXT,
  published_at TEXT,
  thumbnail_url TEXT,
  duration TEXT,
  display_url TEXT
);

CREATE TABLE scraped_sites (
  id INTEGER PRIMARY KEY AUTOINCREMENT,
  run_id INTEGER NOT NULL,
  seed_url TEXT NOT NULL,
  duration_ms INTEGER,
  page_limit INTEGER,
  error TEXT
);

CREATE TABLE scraped_pages (
  id INTEGER PRIMARY KEY AUTOINCREMENT,
  site_id INTEGER NOT NULL,
  url TEXT NOT NULL,
  status_code INTEGER,
  content TEXT
);

Implementation Notes

• Use rusqlite with the bundled feature for portability (zero system dependency on libsqlite3).
• DB writes can be synchronous (rusqlite is sync); wrap in tokio::task::spawn_blocking if called from async context, or keep it simple since writes are tiny.
• For --provider all, create one search_runs row per successful provider response, then one shared row for scraped data if --scrape is active.
• Consider adding DbWriter in a new src/cli/db.rs module.

Acceptance Criteria

• cargo run -- "query" --db test.db creates test.db and populates search_runs + search_results.
• cargo run -- "query" --db test.db --scrape also populates scraped_sites + scraped_pages.
• Running with --provider all stores results from each provider under distinct rows.
• just check passes (fmt, clippy, tests, docs build).
• Unit test in src/cli/args.rs validates --db parses correctly.
• No domain-layer imports of rusqlite or crate::cli::db.

Related Issues

None yet.