v1.3.1 -- AI integration + reader UX overhaul

New features

AI integration

• New Settings → AI integration screen. Bring your own Gemini API key (free tier covers casual use). Stored encrypted in Android Keystore — never persisted to Hive.
• ✨ icon in both novel and manga reader top bars. Tap → AI chapter summary sheet.
• Novel: sends chapter text. Manga / manhwa: fetches each page and sends as inline multimodal payload (sparse-sampled to 25 evenly-spaced pages on very long chapters so it stays under Gemini's request size limit).
• Result cache per chapter — re-asking is instant and doesn't burn API quota.
• Model picker: Gemini 3.5 Flash (default, free), Gemini 2.5 Flash (fallback), Gemini 2.5 Pro (paid, higher quality).

Reader UX

• New floating mini-player pill in the novel reader — capsule with transport buttons, paragraph progress bar, chapter title, expand chevron, and close ×. Show floating Read aloud button toggle in reader settings if you don't want the start FAB.
• Back-to-TTS-Location pill appears at the top when the active paragraph scrolls out of view. Tap = jump back.
• Swipe-up-past-end gesture: at the bottom of a chapter, keep swiping up. After ~120 px of accumulated overscroll the next chapter loads automatically.
• Chapter title is now spoken at the top of each fresh chapter (auto-advance + manual nav) so audio playback has clear chapter boundaries.
• Manual chapter navigation now correctly reloads TTS with the new body text — was reading the previous chapter on tap.
• Speed slider interrupts and restarts the current paragraph mid-utterance so the new rate is audible within a beat instead of after the paragraph ends.
• Chapter-scroll progress bar in the navbar hides while TTS is active (avoided two stacked progress bars at the bottom of the screen).

Neural TTS

• sherpa-onnx Piper voices as a swappable engine alongside flutter_tts. Six voices included in the catalog (Amy, Lessac, Joe, LibriTTS, Alan, Alba). Download manager at Settings → Reading → TTS → Manage voices.