chore(nix): Update refs to submodules & explain why this is needed

[bew]

I saw that you updated harfbuzz' version in ec438e714c25f24783e2a23df5d3fb8eef192b55, and noticed last time that the Nix code currently needs to get the submodules manually, so here you go. I added some explanation why this whole submodule duplication thing is needed at the moment.

I also preferred to use tags instead of commit hash to be much more readable and maintainable.

cc: @happenslol @gabyx is that okay for you?

[happenslol]

Looks good to me, I went through the threads you linked when I created the flake 😅

Linking the tags instead of commit hashes makes sense for readability. I'm worried that the manual upgrade process becomes harder/more tedious though, since now we need to look through the repos and find the tag instead of just using the ref the submodule points to.

[bew]

I went through the threads you linked when I created the flake 😅

yeah I figured, now it's explicit and we can subscribe to something to remove this 'hack' some time in the future (lazy trees is actively being worked on atm, according to recent nix team meeting summaries 👌)

I'm worried that the manual upgrade process becomes harder/more tedious though, since now we need to look through the repos and find the tag instead of just using the ref the submodule points to.

Well, running git submodule status seems to show the result of git describe for each submodules, so it helps a lot:

---

Checking the submodules hashes do match the flake inputs could be quite easily auto-checked by comparing submodules hashes & the input hashes in flake.lock 🤔

[wez]

Thanks!