[DRAFT] initial refactor to make integration-e2e run

[santos1709]

This PR:

• Bumps docker-compose format to 3.8
  • replaces volumes_from with a named bundle_cache volume
  • adds working_dir, deploy.resources memory limit (2500m) and bundle_cache volume
  • removes separate busybox "bundle" service
• Adds JRuby (JR) support for in-memory SQLite in tests
  • require and use activerecord + activerecord-jdbcsqlite3-adapter on JRuby
  • adapt InMemoryAds to use ActiveRecord JDBC connection and convert results to hashes
• Improves S3 support and test fixtures
  • allow per-bucket S3 endpoint format (https://%s.s3.amazonaws.com)
  • pass s3_session_token into Aws::S3::Resource and into returned metadata
  • add s3_session_token to test environment secrets
• Increases YAML parser limit for REST connection
  • sets Psych::Parser.code_point_limit = 10_000_000 before connect
• Updates VCR config to ignore requests to s3.amazonaws.com (and keep previous ignore flag)
  • changes ignore_request block to accept the request and check URI
• Bumps development dependency webmock to ~> 3.26.0
• Moves sqlite3 dev dependency to be platform-aware
  • use activerecord-jdbcsqlite3-adapter + activerecord on JRuby, otherwise sqlite3