[META] Auto undeploy ML Model with TTL

[ylwu-amzn]

Is your feature request related to a problem?
As we continue to develop and utilize numerous Machine Learning models, management and resource optimization have become essential. In particular, we need a mechanism to automatically undeploy the models after they've reached a specific Time To Live (TTL). This TTL limit proposes undeploying models that have been inactive or unused for a particular defined period.

What solution would you like?
Add model TTL to deploy setting. Automatic tracking of model usage or last accessed timestamp to understand if a model should be undeployed. Auto-undeployment should clean up all associated resources to conserve server memory, disk space, or any other resources the models may be using.

What alternatives have you considered?
Keep the same with current user experience. User need to manually undeploy model

Do you have any additional context?
No

[Zhangxunmt]

opensearch-project/documentation-website#7083