finish making bin/speech_to_text_watcher a service managed by systemd and capistrano

[jmartin-sul]

follow on work from #1388, see #1388 (comment)

TODO, in order (question: should this all be completed on QA and stage before we dig into the same work for prod? or should we do prod in parallel with other envs?):

• confirm shared_configs has necessary values for prod, stage, and QA.
  • work with ops to provision the prod AWS resources and credentials. which will require getting the initial stage/QA terraform PR reviewed and merged, a follow-on terraform PR to build the same for prod, and a terraform apply of that infrastructure definition to prod, by ops, since devs only have permission to modify the AWS staging and development envs. Per Jon from ops re: terraform issues: "We're not usually monitoring issues [in the terraform repo] and I don't think people saw [the issues associated with this PR]. Normally all our issues are tracked in [ops tasks]. When you have the PR ready and working, or have questions about getting it working, either file a ticket there or send a message to #ops-aws in slack."
• puppet PR for prod, follow on to https://github.com/sul-dlss/puppet/pull/11366
• implement the capistrano config that allows remote restart and restart on deploy for the STT watcher, similar to what's done for abbyy_watcher

relevant common-accessioning code for capistrano setup:

• common-accessioning/config/deploy.rb

  Lines 46 to 47 in 74a2910

  	# restart abbyy watcher systemd service each deploy
  	before 'deploy:publishing', 'abbyy_watcher_systemd:restart'

• common-accessioning/config/deploy/stage.rb

  Line 13 in 74a2910

  set :abbyy_watcher_server, primary(:worker)

  (and similar for QA and prod)

• common-accessioning/lib/capistrano/tasks/abbyy_watcher_systemd.cap

  Lines 4 to 11 in 74a2910

  	desc 'Restart the abbyy_watcher systemd service'
  	task restart: ['controlmaster:setup', 'otk:generate'] do
  	on fetch(:abbyy_watcher_server) do
  	within release_path do
  	execute 'sudo systemctl restart abbyy_watcher'
  	end
  	end
  	end

totally reasonable to break any of these bullet points out into separate tickets, just link to them here so we can coordinate and avoid duplication of the same work.

see also https://github.com/sul-dlss/terraform-aws/issues/1170 and https://github.com/sul-dlss/terraform-aws/issues/1171

[jmartin-sul]

i'm happy to pick this up, since i've started in on the terraform and puppet stuff, since people often don't want to pick up WIP leftovers. also happy to have someone else run with all or part this if they want to.