Use mime types from server spec for language detection

[bollwyvl]

Elevator Pitch

If available, use the mime types provided by specs to detect languages rather than (just) string parsing or lookup tables.

Motivation

#189 points out a case where a language doesn't (yet) have a well-defined language type (robots-from-jupyter/robotkernel#40).

Design Ideas

In addition to the lookup table of languages, detection should also consult the server spec mime types.

It may be that we need to open this up even wider, and allow for server specs to offer the full rosetta stone of mimetype, codemirror_mode and name, depending on how much information might be available.

[krassowski]

Agreed, this the right thing to do. However, we need to adjust the schema slightly as we need mimetype → language mapping, as one server can handle multiple languages.

[krassowski]

I believe we already do that, so closing as completed. Any future enhancements to the logic would need to be done in @jupyterlab/lsp package.