Improve handling of not existing segments in mesh file

[philippotto]

Since mesh files can be precomputed with a dust threshold, not all segments are guaranteed to be included in it. Therefore, it can happen that the user (indirectly) loads a mesh for such a small segment and then an error message is shown.

Currently, there are multiple problems related to this:

1. the error message is very unspecific. we currently don't know whether there is another bug (e.g., related to proofreading) or if the dust-threshold is at fault.
2. the user is left with this error message and it's hard to know what to do know. if the user directly selected a segment and tried to load that mesh, they can simply use ad-hoc meshing afterwards (the error message will tell them so). however, if the mesh loading was indirectly initiated (e.g., proofreading), this is very hard to fix manually. additionally, the adhoc meshing typically needs mag 1 precision. otherwise, the small segment might not be loaded again.

Ideas:

1. I'm not sure how we can improve this. enumerating skipped segment ids in the mesh file is probably not the way to go, because there will be too many?
2. the front-end could automatically fall back to ad-hoc meshing when a precomputed mesh fails. this should be relatively easy to do.

Example:
https://master.webknossos.xyz/datasets/sample_organization/l4dense_motta_et_al_dev/view#2470,4685,1657,0,0.082

[philippotto]

this should be relatively easy to do.

on a second thought, this won't be straightforward for the proofreading context, as we enrich the mesh objects with mapping-related contexts and this is only implemented for precomputed mesh now.