feature: add telemetry attribution module for SDK usage provenance

[ryantanaka]

Issue #, if available:
N/A

Description of changes:

Adds a new telemetry/attribution.py module to sagemaker-core that provides a structured way to
attribute SDK usage to external integrations such as agent plugins. This attribution if provided will be included in UserAgent header and telemetry requests.

Example Usage:

# other imports
from sagemaker.core import set_attribution, Attribution

set_attribution(Attribution.SAGEMAKER_AGENT_PLUGIN)

# subsequent PySDK code

Adds mechanism to include ARN of resource created for target resources in telemetry requests.

Changes:

• Added sagemaker/core/telemetry/attribution.py with an Attribution enum and set_attribution()
  function. Calling set_attribution(Attribution.SAGEMAKER_AGENT_PLUGIN) sets the SAGEMAKER_PYSDK_ CREATED_BY environment variable, which is picked up by the telemetry emitter and appended as x- createdBy in the telemetry request URL.
• Updated telemetry_logging.py to import _CREATED_BY_ENV_VAR from attribution.py as the single
  source of truth for the environment variable name.
• Re-exported Attribution and set_attribution from sagemaker.core for a clean public import path.
• Added unit tests for the new module in tests/unit/telemetry/test_attribution.py.
• Fixed a pre-existing test bug in test_telemetry_logging.py where feature codes [1, 2] were stale
  and should have been [11, 12] following the update in fix: Update telemetry constants to include MLOPS as well #5438.'
• Added createdBy to user agent metadata.
• Added x-resourceArn to telemetry request URL. This currently applies when TrainingJobs are created, however the implementation allows us to add additional PySDK return types, for which we want to extract a field to be used to set x-resourceArn.
• Fixed bug where telemetry requests were being skipped for calls to *Trainer.train() despite having the telemetry decorator specifically when sagemaker_session is not passed to the constructor of *Trainer(...).