Autoforge

We have optional tensorboard support.
We also now calculate the loss in cielab space using huber loss. In my tests that gave much better color saturation results.
(We also now crank the saturation by 10% in the background to give the loss function a bit of leeway in the colors but even without it the result looks a lot better using the loss changes)

With this my old nemesis the nature.jpg image actually looks a lot better and the greens in the grass is finally visible:

A gui is now finally underway using streamlit. Sadly, this could still take a bit as I have to rewrite a lot of stuff.

Thanks again for rparrett for the tensorboard addition!

What's Changed
* Add optional TensorBoard output by rparrett in https://github.com/hvoss-techfak/AutoForge/pull/16


**Full Changelog**: https://github.com/hvoss-techfak/AutoForge/compare/v1.3.5...v1.4.0

Turns out I missed the multiplicator in the production code for the transmission distance. My bad. This fixes most color problems...

**Full Changelog**: https://github.com/hvoss-techfak/AutoForge/compare/v1.3.4...v1.3.5

**Full Changelog**: https://github.com/hvoss-techfak/AutoForge/compare/v1.3.3...v1.3.4

Reverted a change that broke layer generation

**Full Changelog**: https://github.com/hvoss-techfak/AutoForge/compare/v1.3.2...v1.3.3

I've now printed all filaments I have, recorded them and ran my tests on it. I still have to print the actual files to compare the result, but the color should now be even closer to the actual print.
The height bug is now also fixed and we got support for the mac mps backend now. Thanks rparrett .

What's Changed
* Add `mps` backend support by rparrett in https://github.com/hvoss-techfak/AutoForge/pull/13


**Full Changelog**: https://github.com/hvoss-techfak/AutoForge/compare/v1.3.1...v1.3.2

Just a small release to fix some parameters and make the generation more robust.
I removed one parameter as it seemed to make results worse.

**Full Changelog**: https://github.com/hvoss-techfak/AutoForge/compare/V1.3.0...v1.3.1