feat: add ja.spoilerplus

[kinboy56]

Adds a source for https://spoilerplus.tv, a Japanese raw manga site. Fix #17

The site runs the same CMS and image infrastructure as ja.mangarawjp (same
POST /api/v1/get/c endpoint, same img-cdn.stackpathcdn.app CDN, same
page-descrambling scheme), but ships a different theme, so only the request and
descrambling logic are shared in spirit — the HTML parsing is written against
this site's markup.

Notes on the implementation

• Descrambling. Each page is shuffled into a square grid. The API returns
  the ordering as hex bytes XORed with the site's own domain string, so
  spoilerplus.tv is the key. The order travels to the image through
  PageContext, and PageImageProcessor reassembles the grid on a canvas of
  the original image's dimensions.
• templates/wpcomics was not used. The theme is NetTruyen-based and its
  default selectors do match, but the template does not forward
  PageImageProcessor, and images here are useless without descrambling.
  Changing the template to support that would risk regressions in the seven
  sources already using it, so the source is written standalone.
• Keys are stored decoded. The site percent-encodes series slugs but leaves
  chapter segments as raw UTF-8, and a raw UTF-8 path is answered with a 404.
  Since encode_uri also escapes %, passing the mixed form straight through
  turns %20 into %2520. Keys are therefore decoded on the way in (to_key)
  and encoded on the way out (url_for), so both href forms collapse to one
  key.
• Publication status is read off the labelled row. The info list gives the
  alternative-titles row the same li.row.status classes as the publication
  status row, so selecting by class alone reads the alternative title as the
  status. The row is picked by its 状態 label instead; a regression test
  covers this. An unrecognised value maps to MangaStatus::Unknown rather than
  falling back to Ongoing.
• Inline script bodies are read with html(). Element::data() returns
  None for every <script> on these pages, and the reader's window.MangaId
  / window.CNumber values are what key the image API.
• contentRating is 2 (adult). The genre list carries permanent Ecchi
  and エロい categories, consistent with the sibling sites on the same CMS.
• Search takes precedence over the sort filter. Search has its own endpoint
  that the ordering paths cannot be applied to, so config.hidesFiltersWhileSearching
  is set and a non-empty query wins over whatever sort value is still stored.

Deliberately left out

• date_uploaded is not set. The site mixes relative (1日前) and absolute
  (2026年07月30日) forms, and only the absolute one is parseable — the same
  reason it is unset in ja.mangarawjp.
• No genre filter. /genre/<name>/ exists, but the list changes on the site's
  side, so it would need DynamicFilters rather than a hardcoded list. Left
  for a follow-up.

Test plan

• cargo fmt
• cargo clippy --release
• cargo test --release (17 passed)
• aidoku package && aidoku verify package.aix
• Verified on device: browsing, ranking, search, details, chapter list and
  descrambled pages.