Don't error on duplicate pagination token

[jdisanti]

Motivation and Context

Some AWS operations, such as CloudWatch's GetLogEvents, respond with duplicate next page tokens so that infinite polling can be done (for use cases such as tailing a log). The SDK was previously erroring out in the paginator for this use case (see awslabs/aws-sdk-rust#620).

This PR adds a stop_on_duplicate_token option to the paginators, and removes the error. This solution is the same as what the Go V2 SDK is doing.

Checklist

• I have updated CHANGELOG.next.toml if I made changes to the smithy-rs codegen or runtime crates
• I have updated CHANGELOG.next.toml if I made changes to the AWS SDK, generated SDK code, or SDK runtime crates

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[jdisanti]

I currently have stop_on_duplicate_token defaulting to true, since it felt wrong to default to infinite looping. Not sure this is correct though. Will cross check what other SDKs are doing.

[github-actions]

A new generated diff is ready to view.

• AWS SDK (ignoring whitespace)
• No codegen difference in the Server Test
• No codegen difference in the Server Test Python

A new doc preview is ready to view.

[jdisanti]

It looks like Java V2 uses a customization to mark operations as ending pagination by returning a duplicate token. Go V2 requires the customer to specify a value for its StopOnDuplicateToken. The Rust version of this requires a default with the current fluent paginator builder design, so given the above, I'm going to keep it defaulting to true.

[Velfi]

We should include docs on when this would be helpful to set to false:

When running status-checking operations or operations that you want to continually send in order to "watch" for the latest data (like AWS CloudWatch's GetLogEvents), set this to false. That way, you can continually call the paginator in order to receive new data as it becomes available.

[Velfi]

It looks like Java V2 uses a customization to mark operations as ending pagination by returning a duplicate token. Go V2 requires the customer to specify a value for its StopOnDuplicateToken. The Rust version of this requires a default with the current fluent paginator builder design, so given the above, I'm going to keep it defaulting to true.

I think this is the right approach to take. Infinite looping is something that people shouldn't be able to accidentally trigger unless the docs are very clear about when it happens. I do think we should add some info on why infinite looping is sometimes helpful and guide people to pass false when relevant to their use case.

[github-actions]

A new generated diff is ready to view.

• AWS SDK (ignoring whitespace)
• No codegen difference in the Server Test
• No codegen difference in the Server Test Python

A new doc preview is ready to view.

[github-actions]

A new generated diff is ready to view.

• AWS SDK (ignoring whitespace)
• No codegen difference in the Server Test
• No codegen difference in the Server Test Python

A new doc preview is ready to view.

[github-actions]

A new generated diff is ready to view.

• AWS SDK (ignoring whitespace)
• No codegen difference in the Server Test
• No codegen difference in the Server Test Python

A new doc preview is ready to view.

[github-actions]

A new generated diff is ready to view.

• AWS SDK (ignoring whitespace)
• No codegen difference in the Server Test
• No codegen difference in the Server Test Python

A new doc preview is ready to view.