refactor(compute-providers): share EC2 runner operations

[edersonbrilhante]

Description

Moves shared EC2 runner operations and types out of the control-plane directory so they can be reused by multiple orchestration providers. RunnerSource is required at the caller boundary, preserving the existing scale-up-lambda and pool-lambda values, and both EC2 launch paths use that source for ghr:created_by.

Introduces an Ec2RunnerClient that owns the durable EC2 and SSM clients. Callers bind request state with forRequest({ signal }), allowing the module-scoped control-plane provider to reuse AWS clients across warm Lambda invocations without sharing cancellation state between invocations. The bound operations are passed through scale-up, scale-down, pool, runner configuration, and dynamic-label launch-template lookup instead of using a dependency bag or parameter callback.

AMI values read from SSM are requested with decryption, validated as EC2 AMI IDs, and never logged.

Test Plan

• Compute-provider tests passed: 13 files, 283 tests.
• Control-plane tests passed: 14 files, 348 tests.
• TypeScript type-checks passed for compute providers and control plane.
• ESLint and Prettier checks passed for both affected projects.
• The control-plane NCC bundle completed successfully.
• git diff --check and pre-commit checks passed.

Related Issues

None.

[github-actions]

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

Scanned Files

None