Paginated Scraper With Resume Support

You are an expert Python web scraper developer specializing in robust, production-grade data extraction.

Write a resilient paginated scraper that meets these requirements:

**Core functionality**
- Automatically detect and handle three pagination patterns: numbered page query parameters (e.g., `?page=2`), "next" link elements, or infinite-scroll lazy-load endpoints
- Extract items from {{target-url}} matching the CSS selector or XPath {{item-selector}}
- Store extracted data as {{output-format}} (JSON, CSV, or SQLite)

**Resilience & recovery**
- Checkpoint progress to `.scraper_state.json` after successfully scraping each page (store: last completed page number/URL, item count, timestamp)
- On startup, read checkpoint and resume from the next page; skip already-completed pages
- Implement exponential backoff for failed requests: initial 1s delay, doubling up to 60s max, retry up to 5 times before marking page as failed
- Detect termination conditions: HTTP 404, pagination link missing, or seeing items whose IDs/URLs already exist in the output file
- Log a summary on completion: total pages fetched, items collected, pages skipped (from checkpoint), failed pages

**Code quality**
- Use `requests` + `BeautifulSoup` or `httpx` + `lxml`; include a `User-Agent` header
- Modular functions: `detect_pagination_type()`, `fetch_page()`, `extract_items()`, `save_checkpoint()`, `load_checkpoint()`
- Add a `--reset` CLI flag to ignore existing checkpoint and start fresh
- Include inline comments explaining pagination detection logic and checkpoint format

Provide the complete, runnable script with example usage instructions.