Use generated artifacts for Ruby BDD tests

[nogates]

Summary

• load language-neutral operation, schema, body, and parameter plans at the request-sending step
• replay HTTP interactions through the generated test server instead of Ruby VCR cassettes
• keep the generated Ruby SDK method as the request execution path
• support nested array item schemas from generated request plans
• distinguish the runner internal raw-template marker from legitimate API fields named template
• normalize Ruby empty-body requests so HTTParty does not introduce an unrecorded form content type
• add a local runner that requires neither datadog-api-spec nor openapi-transformer at test runtime

Compatibility usage

The transformer compatibility workflow checks out Ruby master and cherry-picks this PR as one adapter commit. It then generates the complete server, recording database, request plans, and feature tree into the SDK checkout before running Cucumber. No spec-repository files are required at SDK test runtime.

Validation

• complete generated replay against current datadog-api-spec master: 1,685 scenarios discovered; 1,675 passed and 10 SDK-tagged scenarios skipped
• 15,383 Cucumber steps passed
• focused coverage for array bodies, legitimate string/array template fields, and empty-body requests
• Ruby and shell syntax checks

Scope

This PR is the thin Ruby adapter. The language-neutral request construction and generated features live in openapi-transformer; the generated replay database and server are test artifacts.