[Documentation Suggestion] Provide a PDF version of the docs

[Al3cLee]

Hello, thanks for this nice package and a very good documentation! I wanted to learn from the docs today and compiled a PDF version of it via DocumenterTypst.jl.

During the process the only change to be made was simply replace the raw html-style images with standard markdown image syntax

![caption](relative path)

where relative path means: for docs/src/man/symmetries.md file which wants to show an image docs/src/man/img/tree-summary.svg, the link appearing in the markdown file should be ![summary](img/tree-summary.svg). Therefore, it should be quite easy to automate the generation of a PDF file as a workflow.

I didn't think this alone was worth a PR, so I am just opening an issue to see whether you're interested in linking to a PDF version in the documentation; the docs are suitable for long-form reading and I've already printed a copy, haha.

The PDF file is around 2.6MB in size, which wouldn't be a burden for the website. It is available at: https://static.wentaoli.xyz/TensorKit.jl.pdf.

[Jutho]

Hi @Al3cLee , thanks for your enthousiasm.

Since the documentation is not yet complete or finalized, and we hope to have some more updates in the near future, I do think it would be good to commit the required changes/fixes as a PR. Then we can easily automatically rebuild the PDF version.

[Al3cLee]

Sure, I'll be happy to do a PR!

---

The markdown syntax does not control image width, so in some cases the output image can take up very much space, even the whole page, resulting in something like this:

which you might not want. I'm not sure whether hard-wiring size information into the svg will work, and if all other methods don't work, one can still inspect the generated .typ file according to the docs of DocumenterTypst.jl to modify image width.

To be specific, if we generate .typ output as instructed by the docs, the docs/build directory will be populated with documenter.typ and TensorKit.typ where the former is a utility template helper, and the latter is a long file which #imports documenter.typ and can render to a PDF. Using the typst previewer, the preview of this long file can take a long time to load, and the preview itself is laggy. So instead of scrolling through the preview for images, one can simply search for image in TensorKit.typ (they are wrapped in #figure environments, so search for image instead of #image) and tune the width parameter to something other than the default 100%. (I'll also put this info, along with how to build the PDF, into the PR note.)

---

The statement class="color-invertible" will also be absent in the markdown syntax, causing vector graphics to no longer be inverted in dark themes. I've confirmed this behavior on my localhost. It is indeed possible to make a branch dedicated to the typst build, while the main branch is still the color-invertible-graphics html build, but that would cause the author of the docs to manually write the same contents twice just because someone might want dark theme on web, personally I think this is not worth it, haha!