Add historical CDX intelligence views - #6

Summary

Introduce capture-level historical intelligence while preserving the existing normalized-URL workflow as the default.

New analysis views

• --history: first/last seen, capture count, status-code and MIME distributions per normalized URL
• --params: historical prevalence per query-parameter name
• --summary: one aggregate historical record per domain
• TXT and JSONL output, with mode-specific filenames
• matching View selector in the optional local web UI

CDX metadata

• add a CaptureRecord model
• request timestamp,statuscode,mimetype,original only for metadata workflows
• preserve the URL-only contract of iter_original_urls()
• reuse the existing auto/blocks/resume pagination strategy
• historical analysis forces collapse=None so capture counts and first/last seen are not distorted by URL-key collapse

Compatibility and limits

• normal wayparam -d ... behavior is unchanged
• filters, normalization, date ranges, CDX filters, concurrency, proxy and rate limiting are reused
• in analysis modes, --max-results is a global cap on accepted capture rows before aggregation

Tests and docs

• parser and metadata iterator tests
• historical aggregation and mocked HTTP runner tests
• analysis file/stdout tests
• local web UI historical streaming tests
• README, MkDocs guide/reference, internals, API docs and manpage updates

No active target crawling/scanning is introduced; all data still comes from the Wayback CDX API.