Migrate SDKs to centralized evaluator testkit and remove duplicate evaluation tests

Overview

PR #344 introduced the flagd-api-testkit with a dedicated evaluator test suite in evaluator/gherkin/.
This issue tracks the migration of all SDKs to use this new testkit and eliminates duplicated evaluation
tests currently scattered across individual SDK repositories.

Current State

• ✅ Evaluator testkit created (see PR #344)
• ❌ SDKs still running duplicate evaluation tests locally
• ❌ Inconsistent test organization across SDKs

Goals

• Migrate all SDKs to use the centralized evaluator testkit from flagd-testbed
• Remove duplicate evaluation test definitions from each SDK
• Keep only SDK-specific integration/provider tests in each repository
• Establish clear boundaries: evaluation tests (testkit) vs. integration tests (SDK repo)

Required Changes Per SDK

For each SDK (Java, Go, JS, .NET, Python, etc.):

• Remove duplicate Gherkin test files related to evaluation
• Remove unnecessary step definitions for evaluation scenarios
• Update test runner to pull evaluator tests from flagd-testbed via git submodule
• Update CI/CD to run both evaluator testkit and integration-only tests
• Document the new test structure in SDK README

Example Structure

SDK Repository
├── tests/integration/           # Only provider/connection/config tests
│   └── gherkin/
├── flagd-testbed/              # Git submodule
│   └── evaluator/gherkin/      # Shared evaluator tests (via testkit)

Benefits

• ✅ Single source of truth for evaluator tests
• ✅ Faster CI/CD (evaluator tests run without Docker where possible)
• ✅ Reduced maintenance burden
• ✅ Consistent evaluation behavior across all SDKs

Out of Scope (for this issue)

• Creating new features or test scenarios
• Changes to the evaluator testkit itself (separate issues/PRs)

Start with PR #344 and the shared evaluator suite in evaluator/gherkin/, then compare each SDK's duplicate Gherkin tests, step definitions, runners, CI configuration, and README. The work is complete when SDK repositories retain only integration/provider tests, consume flagd-testbed through the submodule, run both test suites in CI, and document the new structure.