fix(high_level): gracefully handle subset_fonts failures

[octo-patch]

Fixes #975

Problem

When doc.subset_fonts(fallback=True) is called during PDF generation, PyMuPDF may raise ValueError: bad 'value' for certain PDFs that have malformed font width tables. This causes the entire translation to crash at the very end — after all translation work has already completed — producing no output file.

Stack trace from the issue:

File "pymupdf/utils.py", line 5848, in subset_fonts
    set_old_widths(font_xref, width_table, def_width)
File "pymupdf/utils.py", line 5617, in set_old_widths
    doc.xref_set_key(df_xref, "DW", dwidths)
ValueError: bad 'value'

Solution

Wrap each subset_fonts() call in a try-except block. If font subsetting fails, log a warning and continue — the output PDF is still written successfully. The only downside is a slightly larger file size when font subsetting is skipped.

Testing

• Verified the fix allows translation to complete when subset_fonts raises a ValueError
• Existing behavior is preserved for PDFs where font subsetting succeeds