Pdf reflow upgrade (#63)

* Replaced PDF-to-Markdown reflow with an enhanced PDF-to-HTML generator.

Specific changes include:
- Replaced `PdfToMarkdownGenerator` and `PdfReflowGenerator` with `PdfToHtmlGenerator`.
- Added native bridge methods in `NativePdfiumBridge.kt` and `pdfium_bridge.cpp` to extract character bounding boxes, page objects, and image pixels.
- Implemented vertical merging of text and images in `PdfToHtmlGenerator` to maintain document layout.
- Added logic to detect and filter repeating headers and footers across PDF pages.
- Updated `ReflowWorker` to generate `.html` files instead of `.md` files and updated `FileType` handling.
- Simplified `SingleFileImporter` by removing dependencies on the legacy Markdown generator.

* Updated PDF to HTML generation to include page breaks and split HTML files into individual chapters based on page markers.

* Added junk character filtering and normalization to PDF text extraction

* - Added `allRecentFiles` to `ReaderScreenState` to track all files, including reflowed versions.
- Updated `recentFiles` in `MainViewModel` to filter out reflowed files (`_reflow`) from the main library view.
- Implemented `deleteBookPermanently` in `MainViewModel` to handle book deletion and cache cleanup.
- Added a "Delete Text View" option to the `EpubReader` controls for reflowed files.
- Improved reflow file detection in `PdfViewerScreen` by checking against `allRecentFiles`.

* Implemented a centralized data-saving mechanism in `PdfViewerScreen` using a debounced `saveAllData` function. This refactor consolidates the saving of annotations, text boxes, highlights, bookmarks, and scroll positions, adding lifecycle-aware triggers and a mutex to ensure data integrity during pauses or document navigation.

* Optimized PDF selection and annotation performance by offloading heavy operations to background threads and improving UI responsiveness.

- Moved PDF page/text opening, character range calculations, and text extraction to `Dispatchers.IO`.
- Wrapped UI updates in `updateSelectionVisuals` and selection logic with `withContext(Dispatchers.Main)`.
- Implemented a more efficient `Popup`-based magnifier to replace manual offsets and transformations.
- Updated `PdfSelectionMenuPopup` properties to prevent focus and click-outside dismissal conflicts.

app/src/main/java/com/aryan/reader/pdf/NativePdfiumBridge.kt

@@ -11,12 +11,19 @@ object NativePdfiumBridge {
     @JvmStatic external fun getPageFontSizes(textPagePtr: Long, count: Int): FloatArray?
     @JvmStatic external fun getPageFontWeights(textPagePtr: Long, count: Int): IntArray?
     @JvmStatic external fun getPageFontFlags(textPagePtr: Long, count: Int): IntArray?
+    @JvmStatic external fun getPageCharBoxes(textPagePtr: Long, count: Int): FloatArray?
 
     @JvmStatic external fun getAnnotCount(pagePtr: Long): Int
     @JvmStatic external fun getAnnotSubtype(pagePtr: Long, index: Int): Int
     @JvmStatic external fun getAnnotRect(pagePtr: Long, index: Int): FloatArray?
     @JvmStatic external fun getAnnotString(pagePtr: Long, index: Int, key: String): String?
 
+    // Image/Object extraction
+    @JvmStatic external fun getPageObjectCount(pagePtr: Long): Int
+    @JvmStatic external fun getPageObjectType(pagePtr: Long, index: Int): Int
+    @JvmStatic external fun getPageObjectBoundingBox(pagePtr: Long, index: Int, outRect: FloatArray): Boolean
+    @JvmStatic external fun extractImagePixels(pagePtr: Long, index: Int, dimens: IntArray): IntArray?
+
     const val ANNOT_TEXT = 1         // Sticky Note
     const val ANNOT_LINK = 2         // Link
     const val ANNOT_HIGHLIGHT = 8    // Highlight