Add per-request limit to parallel remote includes

What does this MR do and why?

After adding the parallel support of remote includes, we needed to limit the number of parallel HTTP requests to avoid abusing IO.

This change adds 2 as the maximum parallel limit per pipeline creation or lint request.

Related to #351250 (closed)

Implementation table: #351250 (comment 1647814064)

Screenshots or screen recordings

1. Enable the feature flag;

   Feature.enable(:ci_parallel_remote_includes)

2. Run a basic web server to test slow includes.

   require 'rack'
   
   def run(env)
     puts "URL: #{env['REQUEST_URI']} started at #{Time.now}"
     sleep 5
     puts "URL: #{env['REQUEST_URI']} finished at #{Time.now}"
     "
     test:
       script: echo 'hello world'
     "
   end
   
   app = Proc.new do |env|
     [
       200,
       { "Content-Type" => "text/x-yaml" },
       [run(env)]
     ]
   end
   
   Rack::Server.start(app: app, Port: 9292)

3. Enable local includes on the Admin page http://gdk.test:3000/admin/application_settings/network
4. Use a basic CI config;

   include:
     - remote: http://0.0.0.0:9292/file1.yml
     - remote: http://0.0.0.0:9292/file2.yml
     - remote: http://0.0.0.0:9292/file3.yml
     - remote: http://0.0.0.0:9292/file4.yml
     - remote: http://0.0.0.0:9292/file5.yml
   
   build:
     script: exit 0

5. Result;

MR acceptance checklist

This checklist encourages us to confirm any changes have been analyzed to reduce risks in quality, performance, reliability, security, and maintainability.

• I have evaluated the MR acceptance checklist for this MR.