The `getSharedChannelsService` method was checking as well for the
Shared Channels to be active, which only the lead node of a cluster
is, so API operations that should run correctly like sharing/unsharing
a channel or inviting/uninviting a remote were returning a 400 bad
request.
This change updates the method to check for the Shared Channel service
to be active only on request, and on doing so it changes the error and
status code returned to indicate specifically that the service is
running but inactive, and returns a 500 as the situation is not an
error on the requester.
Co-authored-by: Mattermost Build <build@mattermost.com>
0704500609