Add loaded meshes to sharing link

[daniel-wer]

• Includes some cleanup (review first commit separately to have an isolated look)
  • I was annoyed that because flow returns mixed for Object.values, we often had to add $FlowIssue comments, or worse, adapted the code to work around that by using Object.keys, converting the key to int and then mapping to get the values. I introduced Utils.values which returns the correct type and should be the only spot where the $FlowIssue comment is needed. (_.values also seems to be correctly typed, but has unknown overhead)
  • A couple of times I specified the type if the values were mapped, afterwards, which is a separate issue, since mapping Array<number> will lead to flow assuming a type of number | any for the mapped value argument 🤷 I did not systematically look for these occurrences, but I assume there are quite a few in our code, leading to less type security.
  • Flow Coverage improved:

    Before │project │ percent │  total │ covered │ uncovered │
           │ oxalis │    88 % │ 145209 │  127907 │     17302 │
    After  │project │ percent │  total │ covered │ uncovered │
           │ oxalis │    88 % │ 145166 │  127953 │     17213 │
    

• I cleaned up the isosurface related code quite a bit. Most prominently, I removed the functionality where flycam position changes triggered a continuation of isosurface loading under some circumstances. The respective code was triggered quite a lot (and usually for nothing) and we decided that it is obscure and not really useful. I also removed some unused code and parameters.
• I converted the precomputed mesh loading code to a saga, so it can be easily triggered by an action call and also easily be canceled if the user deletes the mesh while it's loading.
• The loaded meshes are now included in the sharing link and are automatically loaded if such a link is opened. I included only the currently visible meshes in the sharing link.
• Sort the precomputed mesh chunks by their distance to the seedPosition before loading them, so that the mesh loads from the seedPosition out.

URL of deployed dev instance (used for testing):

• https://meshesinlink.webknossos.xyz

Steps to test:

• Test mesh related functionality (both in a tracing as well as in view mode!)
  • Load ad-hoc/precomputed meshes
  • Download meshes
  • Reload ad-hoc meshes
  • Clicking on mesh entries should jump to the seed position
• Test that meshes can be shared via link
  • Load a couple of meshes (ad-hoc and precomputed)
  • Copy the sharing link from the share modal or by clicking the button right of the position view
  • Open a new tab and enter the url -> The meshes that were visible before should be automatically loaded
• Test that mesh loading is aborted if the mesh is deleted while loading
  • Open the network tab and throttle the network speed
  • Load meshes and delete them while loading
• Test refreshIsosurfaces api method
  • Open a volume tracing without a fallback layer
  • Brush 2 different segments (4 slices, otherwise you might not get a mesh!)
  • Compute ad-hoc meshes for both of them
  • Modifiy one of the segments
  • Call window.webknossos.apiReady(3).then(api => api.data.refreshIsosurfaces());
  • -> Only the modified isosurface should be reloaded

Issues:

• contributes to Add layer configuration to sharing link #5963

---

(Please delete unneeded items, merge only when none are left open)

• Updated (unreleased) changelog
• Updated documentation if applicable
• Ready for review

[daniel-wer]

@philippotto I have mostly tested the happy paths for this PR, but will do some more monkey testing next week. Still, this should be ready for a first review :)

[philippotto]

Great work & refactoring 👍 It's a joy to read through these changes 🕺

I only have some concerns with processTaskWithPool. Maybe they are unfounded (I didn't test my assumption), but my current understanding is that this needs some changes.. Happy to call about this, if you want 🤙

[philippotto]

Great adaptation of the pool "saga" 🥇 I couldn't get the sharing link to work, yet, though. I tried it with adhoc meshes and the URL doesn't seem to contain anything mesh related 🤔

[philippotto]

I wasn't sure, whether the new round is ready for review again, but especially the last commit seemed like a bigger chunk ready to review :) Great work refactoring this! I only left two suggestions. Let me know when/whether I should also test again :)

[daniel-wer]

I wasn't sure, whether the new round is ready for review again, but especially the last commit seemed like a bigger chunk ready to review :)

Thanks a lot!

I addressed your feedback, would be great if you could do another test round.
I'll also test some more and will add a screenshot test as well.

[philippotto]

Sorry, this PR slipped through on my side! I reviewed the latest commits and also did another testing round. Looks very solid 👍 Looking forward to see this live 🎉

[daniel-wer]

Nightly test ran through well: https://app.circleci.com/pipelines/github/scalableminds/webknossos/5717/workflows/22bbe067-6ff4-4a38-a907-c37a4132ee19/jobs/17004

[philippotto]

Wonderful! Thank you so much for taking care of the screenshot tests 🙏 I reviewed the latest commits and everything looks good!

[philippotto]

Has this become necessary due to the update to pixelmatch? Or has this always been an issue? Good catch, either way 👍

[daniel-wer]

From what I read and experienced, pixelmatch started to assert this in the new version but the problem existed before. Usually the difference was fairly small (which is why it mostly worked before), but I also encountered garbled diffs during my testing with other --use-gl flags.