feat: add JavBus Python scraper (supports censored + uncensored search)

[xy952666680]

Summary

New JavBus Python scraper supporting both censored and uncensored search.

• sceneByFragment / sceneByQueryFragment / sceneByName / sceneByURL
• Searches javbus.com/search/<code> first, falls back to /uncensored/search/<code> when no censored results
• Hardcoded Cookie: over18=18; existmag=mag to bypass the site's age gate
• Stdlib-only (urllib + regex), no third-party dependencies

Scraper types

• sceneByFragment
• sceneByName
• sceneByQueryFragment
• sceneByURL

Examples to test

• Censored: SSIS-123 → https://www.javbus.com/SSIS-123
• Uncensored: HEYZO-3016 → https://www.javbus.com/HEYZO-3016

Notes

• Code extraction strips junk tokens (CD1, DISC2, JAV24, ...), supports alphanumeric prefixes (s2mbd-055 → S2MBD-055), normalizes dashless codes (DDT246 → DDT-246), and ignores dates before the code
• Fragment paths return a single scene object; sceneByName returns a list
• Duration returned in seconds (Stash convention)

[xy952666680]

Review fixes applied (verified against live javbus.com)

This branch includes fixes from a full review — Codex CLI review plus end-to-end testing against live javbus.com pages (censored + uncensored), both locally and on a staging Stash instance.

1. Fragment dispatch: sceneByFragment / sceneByQueryFragment now read code/title/urls (Stash scene fragments have no name field), return a single scene object (Stash unmarshals fragment paths into one scene — see script.go), and prefer an exact code match over the first fuzzy search hit.
2. Duration: converted minutes → seconds (×60); key omitted when the page has no runtime.
3. Code extraction ec(): strips CD|DVD|DISC|PART|JAV junk tokens, supports alphanumeric prefixes (s2mbd-055 → S2MBD-055), normalizes dashless codes (DDT246 → DDT-246), and is immune to dates appearing before the code.
4. Tags: regex anchored to class="genre"><label> so performer names no longer leak into the tag list.
5. SSL: ssl.create_default_context() instead of an unverified context.
6. YML: removed the dead validSearch argument (the script never read it; Stash doesn't pass query strings to script scrapers).

Tested entry points (all against live pages):

• Censored SSIS-123 and uncensored HEYZO-3016 × fragment (code-only) / name / url
• Edge cases: junk-token codes (JAV24 CD1 ABC-123 → ABC-123), dashless codes, empty fragment → {}
• Fields verified: title, code, date, duration (seconds), director, studio, tags (no performers), performers, image