♻️ Refactor PDF processing code to align with PDFium thread-safety and rendering best practices

[balazs-szucs]

• Resolves concurrent rendering crashes by serializing PDFium FFI calls with a global mutex lock.
• Moves blocking page analysis tasks to a dedicated thread pool to prevent problems due to newly added Locks.
• Improves low-resolution rendering by using configured DPI settings and preserving aspect ratios.
• Refactors PDF stuff to use PDFium-render, and removes now (unused) pdf.rs library.

On the last point, generally, PDFIum should be much more reliable/performant than from what I can moreso a hobbyist Rust PDF library. So, using PDF.rs then to fallback PDFium, that seemed like bit of strange choice, and when I flipped them, i thought, there is very unlikely that PDFium would fail and PDF.rs would catch it, so at that point better to just remove. PDF.rs was also pulling in seemingly quite large of other deps. So, I think that's a win.

Date parsing still hacky. But it is mostly in working condition.

I am still not happy with date stuff, but otherwise this should be improvements. It's able parse in the bit more metadata (e.g., Summary/Description that was previously ignored), i have not been able to trigger segfault, but yeah bit slower overall (not a huge surprise.)