[azure-ai-ml] Re-record three e2e tests blocked by current live resources

[Chakradhar886]

Three e2e tests cannot currently produce valid playback recordings while removing temporary sanitizer-era live-only markers:

• component/e2etests/test_component.py::TestComponent::test_load_component_from_flow_in_registry: live testFeed create returns HTTP 400 UserError because FlowDefinitionDataUri on the uploaded flow blob has an invalid format.
• environment/e2etests/test_environment.py::TestEnvironment::test_environment_create_or_update_docker_context: all accessible succeeded SDK workspaces have backing storage with shared-key access disabled; upload fails with KeyBasedAuthenticationNotPermitted.
• internal/e2etests/test_pipeline_job.py::TestPipelineJob::test_pipeline_with_setting_node_output_mode: setup uploads components to workspace storage and fails with the same KeyBasedAuthenticationNotPermitted policy before job submission.

Required follow-up: repair the registry flow service path and provide a recording workspace compatible with artifact uploads, then run with AZURE_TEST_RUN_LIVE=true, verify sanitizer output, publish the cassettes, remove the linked live_test_only markers, and confirm playback.

Additional verified recording-workspace blockers:

• dsl/e2etests/test_dsl_pipeline_samples.py::TestDSLPipelineSamples::test_parallel_components_with_tabular_input_pipeline_output: live validation fails because the configured workspace does not contain the required cpu-cluster compute.
• dsl/e2etests/test_dsl_pipeline_samples.py::TestDSLPipelineSamples::test_parallel_components: live artifact upload receives a non-HTTPS datastore endpoint and Azure Core rejects bearer authentication.