fix(mangadar): support new rows format and iterate all x-data containers

[hamedhani1998]

Summary

• Iterate all div[x-data] containers to find chapter data instead of single selectFirstn- Support new compact array format: ows: [[id, number, url, timestamp], ...]n- Keep legacy chapters: [{...}] format as fallback
• Add parseRowsJson helper for compact array parsing

[Starmania]

Please keep the legacy chapters: branch only if you can point at a title that still uses it

MangaDar.kt:221-231, and Dto.kt:22-36

I couldn't find the old format anywhere. Sampled 7 titles spanning 2 to 541 chapters — 877 rows in total:

title	chapters	format
all-hail-the-sect-leader	541	rows:
blue-box	247	rows:
do-your-best-to-regret	40	rows:
hands-off-my-cleaner	35	rows:
green-skin	10	rows:
bungou-stray-dogs-dazai-wo-hirotta-hi	2	rows:
when-love-calls-upon-me	2	rows:

Every one of those pages has exactly 10 div[x-data], with rows: always in container index 7 and every row of length 5. A \bchapters: search matches nowhere on any of the seven pages.

That makes the legacy branch unreachable, and ChapterDto unreachable with it — Dto.kt:22-36 has exactly one consumer, the parseAs<List<ChapterDto>>() call at :228.

Seven titles is a sample, not a proof, so: if you know of a title that still serves chapters:, name it and the fallback is justified — keep it. Otherwise please delete the legacy branch and ChapterDto along with it. A fallback for a shape that can't be reproduced is speculative, and it's the kind of code that quietly rots. (SearchResponse / SearchMangaItemDto are unaffected.)