Logic to control Pylance auto-indent experiment

[debonte]

Updated NodeLanguageServerAnalysisOptions to pass a new autoIndentSupport initialization option to Pylance. This is set to true if the user is in the "pylanceAutoIndent" experiment and editor.formatOnType is enabled.

Additionally, if the user is in the experiment and formatOnType is not set in the Python-specific config settings, we will set it to true. Users can opt out of the experiment by setting it to false in the [python] settings, like this:

"[python]": {
    "editor.formatOnType": false
}

If the user is not in the experiment, we will unset the Python-specific formatOnType setting (if present), unless it is set to false. I was concerned that users might opt-out and then get removed from and added back into the experiment. If that happens, I don't want them to need to opt out again. Happy to remove this logic if this scenario won't happen.

Since this code is temporary, I chose not to integrate it into PythonSettings in any way. Assuming we decide to enable this feature for everyone at some point, we'll remove this code and change the default of editor.formatOnType for Pylance users to true via Pylance's package.json:

"configurationDefaults": {
    "[python]": {
        "editor.formatOnType": true
    }
},

Users running internal (-dev) builds of Pylance will be treated as bring in the experiment treatment group. Otherwise we would always unset editor.formatOnType for them which would make testing awkward.

Also includes a related change in LanguageClientMiddleware to return the Python view of editor.formatOnType so Pylance can include it in its settings telemetry event.