[fix] use engine-type when looking up supported_languages from JSON files

searx/data/engines_languages.json stores language information for
several searchengines in a json endoded dict that maps engine-"types" to
their supported languages; for instance there is an entry "google",
mapping to the supported languages of the google engine.

However, the lookup code did not use the engine 'type' (as in: the
filename searx/engines/<enginetype>.py), but instead the manually
configured engine name from settings.yml when querying. This is
problematic as soon as users start to specify additional engine
instances with custom names in the config file, as for instance
suggested as a workaround for multilingual search in the manual[0]:

> engines:
>   - name : google english
>     engine : google
>     language : english

Here, the engine name "google english" will be used for the lookup in
the json file, which does not exist. The empty supported_languages then
lead to a type error later in the processing callchain.

This patch changes the behaviour to use the engine's entry-"type"
("google" in the above example) for the lookup. This should fix bug #2928.

0: https://searx.github.io/searx/user/search_syntax.html#multilingual-search