Allow using existing guid for run_single_worker

When running single workers in the context of a worker monitor, the
worker records are usually already created by the monitor process by the
time the worker is forked/spawned.

Allowing the lib/worker/bin/run_single_worker.rb script to pass a guid
allows it to use those records instead of creating new ones.  If this
option isn't used, the already created ones exist along side the other
worker records, and cause evm:stop to not function properly.

lib/workers/bin/run_single_worker.rb

@@ -23,6 +23,10 @@
     options[:dry_run] = val
   end
 
+  opts.on("-g=GUID", "--guid=GUID", "Find an existing worker record instead of creating") do |val|
+    options[:guid] = val
+  end
+
   opts.on("-h", "--help", "Displays this help") do
     puts opts
     exit
@@ -52,7 +56,12 @@
 worker_class = worker_class.constantize
 worker_class.before_fork
 unless options[:dry_run]
-  worker = worker_class.create_worker_record
+  worker = if options[:guid]
+             worker_class.find_by!(:guid => options[:guid])
+           else
+             worker_class.create_worker_record
+           end
+
   begin
     worker.class::Runner.start_worker(:guid => worker.guid)
   ensure