Revert to use only a single worker by default on outputs

The move to auto-scale output workers was great in theory, but a lot of outputs
just weren't built to support it well, they often used too many resources or had
logical errors.

Fixes #4905

logstash-core/lib/logstash/pipeline.rb

@@ -19,6 +19,8 @@ require "logstash/output_delegator"
 module LogStash; class Pipeline
   attr_reader :inputs, :filters, :outputs, :worker_threads, :events_consumed, :events_filtered, :reporter, :pipeline_id, :logger, :thread, :config_str, :original_settings
 
+  DEFAULT_OUTPUT_WORKERS = 1
+
   DEFAULT_SETTINGS = {
     :default_pipeline_workers => LogStash::Config::CpuCoreStrategy.maximum,
     :pipeline_batch_size => 125,
@@ -413,16 +415,12 @@ module LogStash; class Pipeline
     klass = LogStash::Plugin.lookup(plugin_type, name)
 
     if plugin_type == "output"
-      LogStash::OutputDelegator.new(@logger, klass, default_output_workers, *args)
+      LogStash::OutputDelegator.new(@logger, klass, DEFAULT_OUTPUT_WORKERS, *args)
     else
       klass.new(*args)
     end
   end
 
-  def default_output_workers
-    @settings[:pipeline_workers] || @settings[:default_pipeline_workers]
-  end
-
   # for backward compatibility in devutils for the rspec helpers, this method is not used
   # in the pipeline anymore.
   def filter(event, &block)

logstash-core/spec/logstash/pipeline_spec.rb

@@ -209,7 +209,7 @@ describe LogStash::Pipeline do
         pipeline.run
 
         expect(pipeline.outputs.size ).to eq(1)
-        expect(pipeline.outputs.first.workers.size ).to eq(pipeline.default_output_workers)
+        expect(pipeline.outputs.first.workers.size ).to eq(::LogStash::Pipeline::DEFAULT_OUTPUT_WORKERS)
         expect(pipeline.outputs.first.workers.first.num_closes ).to eq(1)
       end