Update AWS SDK v2 to 2.20.89

[wendigo]

No description provided.

[findinpath]

@trinodb/maintainers pls run the PR with secrets

[losipiuk]

This was here for a reason

[losipiuk]

cc: @linzebing

[losipiuk]

Now when you changed fromByteBufferUnsafe it should be fine.
@linzebing PTAL.

[losipiuk]

Restore ByteBufferAsyncRequestBody.

BTW: Even if we wanted to drop it - it should be separate commit - not SDK version update

[wendigo]

It was introduced in aws/aws-sdk-java-v2#3925 (https://github.com/aws/aws-sdk-java-v2/releases/tag/2.20.84)

[wendigo]

@losipiuk extracted to separate commit

[wendigo]

@losipiuk I think this part is relevant:

The static methods in AsyncRequestBody do some defensive copying and transforming of types that hamper performance and utilise more memory than needed if the user has already ensured either that concurrent modification won't happen or that it doesn't matter, this pr addresses that with new AsyncRequestBody constructors.

This also fixes an unnecessary copy being made in AsyncRequestBody#fromByteBuffer (https://github.com/aws/aws-sdk-java-v2/issues/1735) and AsyncRequestBody#fromString as a side effect.

[wendigo]

There is also pending PR which seems relevant here: aws/aws-sdk-java-v2#4096:

This change is effectively running in my own production code where I upload files via [MappedByteBuffer](https://docs.oracle.com/javase/8/docs/api/java/nio/MappedByteBuffer.html), when performance testing I was able to achieve a 25% performance improvement over TransferManager uploading a 1GB file.

[losipiuk]

/test-with-secrets 7d308de

issue removed

[losipiuk]

/test-with-secrets sha=7d308deed56c6cc619d16b1b7a18a7713972e586

[electrum]

Looks like we need to squash the commits, since the internal class ByteArrayAsyncRequestBody is removed in the new version, and thus the first commit won't compile on its own.