Fix invalid terraform drifting to redis in global replication groups

[jramosf]

Description

This change follows-up on #39745 and fixes a bug which causes drift for ElastiCache Global Datastore in Valkey.

This is due to the current hard-coded default value for the engine argument, which inevitably causes the drift even in a terraform plan in an empty state importing an existing aws_elasticache_replication_group resource with the global_replication_group_id argument set (if this argument is set, the redis value in engine argument should not be passed).

      ~ engine                         = "valkey" -> "redis" # forces replacement

The API parameter in AWS for Engine is not required and has a default value, which brings the Terraform provider in line with AWS API behaviour.

Relations

Relates #39745

References

Output from Acceptance Testing

As per the contributor guide I would prefer the tests run on HashiCorp side.

[github-actions]

Community Note

Voting for Prioritization

• Please vote on this pull request by adding a 👍 reaction to the original post to help the community and maintainers prioritize this pull request.
• Please see our prioritization guide for information on how we prioritize.
• Please do not leave "+1" or other comments that do not add relevant new information or questions, they generate extra noise for issue followers and do not help prioritize the request.

For Submitters

• Review the contribution guide relating to the type of change you are making to ensure all of the necessary steps have been taken.
• For new resources and data sources, use skaff to generate scaffolding with comments detailing common expectations.
• Whether or not the branch has been rebased will not impact prioritization, but doing so is always a welcome surprise.