Update service-2.json with latest public botocore service model

[lucasjia-aws]

Summary

Replaced the bundled service-2.json with the latest public botocore SageMaker service model and updated codegen tooling + downstream packages to accommodate differences in the new model.

Changes

File / Area	Change	Why
sample/sagemaker/2017-07-24/service-2.json	Replaced with latest public botocore service model	Core change — keeps the bundled model in sync with the public botocore release
tools/constants.py	Added PIPE_VAR_OVERRIDES for ResourceConfig fields; extended REQUIRED_TO_OPTIONAL_OVERRIDES for ModelPackage/SecurityConfig; added IntPipeVar mapping; added SHAPES_CODEGEN_OUTPUT_DIR	New model exposes fields (e.g. InstanceCount, VolumeSizeInGB) as plain integers that need to be pipeline-variable-aware; some members marked required in the model are optional in practice
tools/resources_extractor.py	Added EXCLUDED_RESOURCES set to exclude Session	New model surfaces a Session resource that collides with boto3.session.Session import
tools/shapes_extractor.py	Integrated PIPE_VAR_OVERRIDES into shape member type resolution	Allows codegen to emit IntPipeVar instead of int for overridden fields
tools/shapes_codegen.py	Added IntPipeVar import in generated code; updated default output dir to shapes/	Generated shapes need the new type; output path aligned with project layout
tools/resources_codegen.py	Added IntPipeVar to PYTHON_TYPES list	Ensures IntPipeVar fields are recognized as configurable attributes
resources.py, shapes.py, config_schema.py, shape_dag.py, resource_plan.csv	Regenerated by codegen	Reflect the updated service model and tooling changes above
sagemaker-mlops/ and sagemaker-train/	Fixed import and compatibility issues	Downstream packages referenced shapes/fields that changed in the regenerated code

Testing

• Codegen runs successfully and produces parseable Python output
• All unit tests pass locally
• Integration tests running in CI