perf: optimize image processing with larger pixel chunks

[CordlessCoder]

I'm seeing a ~2x speedup thanks to this :)

[CordlessCoder]

Bumped dependency versions and updated image features to work with the new defaults.

Didn't find any performance regressions when running with .jpeg inputs.

[doprz]

Thank you for another PR @CordlessCoder , it is very much appreciated!

[CordlessCoder]

The PR is ready to merge btw.

[doprz]

LGTM!

Thank you once again for another PR @CordlessCoder !

Here are some of my thoughts:

• Increasing the chunk size from 4 to 4096 should significantly improve performance by:
  • Enhancing cache locality
  • Reducing threading overhead per pixel
  • Allowing better SIMD optimization opportunities (in the future)
• How does this handle images whose dimensions aren't evenly divisible by the chunk size? The chunks_exact_mut will process complete chunks, but we might need to handle remaining pixels.

[CordlessCoder]

LGTM!

Thank you once again for another PR @CordlessCoder !

Here are some of my thoughts:

* Increasing the chunk size from 4 to 4096 should significantly improve performance by:
  
  * Enhancing cache locality
  * Reducing threading overhead per pixel
  * Allowing better SIMD optimization opportunities (in the future)

* How does this handle images whose dimensions aren't evenly divisible by the chunk size? The `chunks_exact_mut` will process complete chunks, but we might need to handle remaining pixels.

There are no remaining pixels, it uses .par_chunks_mut - not .par_chunks_exact_mut.

[doprz]

Thank you for clarifying that; I will be updating the docs, nix flake, and possibly finish #25 for the v1.1.0 release soon.