Change dict loading mechanism from internal static to server extension

[krassowski]

Builds upon #70

Closes #31

• dictionaries now have an ID allowing to distinguish multiple dictionaries from different locations for the same language/code
• OS-specific locations are now searched to discover pre-installed dictionaries
• use babel to resolve the language/locale name for given locale code (as it is already a dependency of JupyterLab) - we can revisit this in the future

In the follow up PR: split the existing dictionaries into separate packages so that one does not need to install the dictionaries they don't need

[github-actions]

👈 Launch a binder notebook on branch krassowski/jupyterlab_spellchecker/ocordes-server-ext

[ocordes]

This looks good! I agree to look for dictionaries manually, without a lang.json file. But how can we implement a language description which is able to load files directly from the network? With lang.json this is easy, okay, one can argue "download everything you need" is sufficient ;-)

[ocordes]

Very good :-)

[krassowski]

But how can we implement a language description which is able to load files directly from the network? With lang.json this is easy, okay, one can argue "download everything you need" is sufficient ;-)

I have mixed feelings here. I did think about making another setting for "online dictionaries" so that one would go to the advanced settings and add something like:

{
    "ignore": ["JupyterLab", "Notebook", "Notatnik"],
    "language": "pl_PL-online",
    "onlineDictionaries": [
        {
            "id": "pl_PL-online",
            "aff": "http://some-url/pl_PL.aff",
            "dic": "http://some-url/pl_PL.dic",
            "name": "polski (Polska)"
        }
    ]
}

I'm not convinced that this is necessary now, given that the dictionaries loaded locally to the disk are more reliable, but I see that for some very specific use cases this might be useful. I will open a new issue to track this one.