Fix cluster integration HTTP adapter

Somehow, https://gitlab.com/gitlab-org/gitlab/-/merge_requests/76879
caused typhoeus to become the default adapter for all elasticsearch
clients, causing problems for the cluster integration, which apparently
has Net::HTTP specific needs.

See https://gitlab.com/gitlab-org/gitlab/-/issues/350519

Changelog: fixed

app/models/clusters/concerns/elasticsearch_client.rb

@@ -15,7 +15,7 @@ module Clusters
 
           proxy_url = kube_client.proxy_url('service', service_name, ELASTICSEARCH_PORT, ELASTICSEARCH_NAMESPACE)
 
-          Elasticsearch::Client.new(url: proxy_url) do |faraday|
+          Elasticsearch::Client.new(url: proxy_url, adapter: :net_http) do |faraday|
             # ensures headers containing auth data are appended to original client options
             faraday.headers.merge!(kube_client.headers)
             # ensure TLS certs are properly verified

spec/support/shared_examples/models/clusters/elastic_stack_client_shared.rb

@@ -47,7 +47,7 @@ RSpec.shared_examples 'cluster-based #elasticsearch_client' do |factory|
       it 'copies proxy_url, options and headers from kube client to elasticsearch_client' do
         expect(Elasticsearch::Client)
           .to(receive(:new))
-          .with(url: a_valid_url)
+          .with(url: a_valid_url, adapter: :net_http)
           .and_call_original
 
         client = subject.elasticsearch_client