Miyo v0.0.4

This release focuses on reader reliability, download integrity, and improved handling for CAPTCHA-protected sources.

Highlights

Improved Cloudflare/CAPTCHA handling for sources that complete verification without exposing a traditional cf_clearance cookie.
Hardened offline reading and page caching to avoid blank, stale, partial, or mismatched pages.
Improved download integrity so incomplete chapters are no longer marked as successfully downloaded.
Cleaned up source deep links and browser source initialization issues found during release validation.

Accessibility-based CAPTCHA session verification for WebView flows.
Safer page cache writes using temporary files before committing final cache entries.
SHA-256 cache file naming to reduce filename collision risk.
Invalid cache eviction before retrying page loads from the network.

Chapters could be marked complete even when one or more pages failed or were skipped during download.
CAPTCHA sessions could fail after a valid solve if the source did not expose a Cloudflare-named cookie.
Concurrent resolver launches could overwrite an active continuation and leave the first request unresolved.
Reader page tasks could be reused incorrectly when different pages shared the same page ID.
Empty, corrupt, or partial cached images could be shown as valid reader pages.
WebView cookie persistence could drop same-host URLs with different paths.
Several source web links had malformed manifest host entries.
Duplicate source deep-link entries were removed.
Browser source initialization now uses the correct source factory, fixing the release CI compile failure.

WebView cookies restored from Android's CookieManager now preserve narrower scope when full cookie metadata is unavailable.
CAPTCHA verification now checks actual page accessibility before accepting or rejecting a session.
Reader cache reuse now validates cached files before returning them to the UI.