mupdf: update to 1.24.0.

List of changes in MuPDF 1.24

Error handling changes:

You must call pdf_report_error in the final fz_catch. Any
unreported errors will be automatically reported when a new
error is raised, or when closing the fitz context.

New formats:

Read Office (XML) files! We internally open and convert
docx/pptx/xlsx documents to HTML to allow reading the plain
text content. The exact layout will NOT be preserved.

Optional compile time option to use libarchive for reading CBR
and other archive formats.

Read plain text documents.

Read gzipped files directly.

Open and read FDF files to support importing annotations or
form data using the low-level PDF functions. There are no tools
for this yet.

Read CFB (Compound File Binary) format archives -- used for the
Office formats.

Write images as JPEG2000.

New tools and features:

mutool bake (and associated functions) to bake appearance of
annotations and forms into static content.

Font subsetting flag to mutool clean (EXPERIMENTAL FEATURE).

Option to use ObjStms when writing PDF files.

Compression effort option when writing PDF files.

Add option to control how line art is affected by redaction.
Add more options to control how images are affected by
redaction (remove-unless-invisible).

Fix up q/Q gstate balance when cleaning content streams.

New functions and types:

pdf_rearrange_pages to subset or re-order pages in a PDF file.

fz_invert_bitmap to invert monochrome bitmaps.

fz_compressed_image_type to query the format of a compressed
image.

fz_text_decoder to convert various legacy and CJK encodings
into UTF-8.

More helper functions to easily manipulate PDF objects in C.

Add flag to control fz_place_story overflow behavior when the
text doesn't fit into the box.

New archive handlers can be added at runtime.

Major bug fixes and improvements:

Support using Art, Bleed, Media, and Trim boxes for PDF page
size.

Support ActualText in PDF! No more strange text extraction when
the file uses ActualText to patch over bad font encodings.

Add special TrueType fallback encoding CMap for a specific
flavor of broken PDF files that use an "identity" encoding
without embedding the font.

Limited "transfer function" support in PDF. Transfer functions
are a deprecated legacy PDF feature that predates proper color
management. They were intended to provide limited color
management such as applying a gamma curve. Transfer functions
have often been (ab)-used to invert images, and many PDF
creators use them when writing softmask images. We have added
support for this case only.

Box drawing characters added to fonts for HTML and plain text
documents.

Write more compact PDF files (removed some unneccessary
whitespace).

Improved selection behavior for non-axis aligned text.

Improved heuristics for detecting the logical and visual order
of RTL text in PDF.

Improved heuristics for inserting missing spaces in PDF text.

Improved handling of CMYK JPEG files (which ones are inverted
and which are not).

Improved content type detection. Don't assume everything is PDF
when we can't recognize it.

Removed deprecated functions:
pdf_check_signature

(wiz)