feat: add AWS Translate resource support

[chrisns]

Summary

• Add three new AWS Translate resource types: TranslateParallelData, TranslateTerminology, and TranslateTextTranslationJob
• Follows the SDK v2 interface + gomock pattern established by the RAM resources
• ParallelData and TextTranslationJob resources include Filter() to skip resources in terminal/non-actionable states
• TextTranslationJob uses StopTextTranslationJob (stop, not delete) since batch jobs cannot be deleted

Details

New resources:

Resource	List API	Remove API	Filter
TranslateParallelData	ListParallelData	DeleteParallelData	Skips DELETING, FAILED
TranslateTerminology	ListTerminologies	DeleteTerminology	None needed
TranslateTextTranslationJob	ListTextTranslationJobs	StopTextTranslationJob	Only SUBMITTED, IN_PROGRESS

Design decisions:

• No tags — would require an extra ListTagsForResource call per resource; can be added later
• No TargetLanguageCodes field — array not useful for filtering
• Shared TranslateAPI interface for all three resources, matching the RAM pattern

Note: Translation jobs can only be stopped, not deleted. The AWS Translate API does not provide a way to remove completed/stopped job metadata — it will remain visible in the account until it ages out naturally. This is an AWS limitation, not something we can address.

[ekristen]

At a glance things look great, have one question on the filter for translation job.

Also I have to ask, have you fully tested these resources against AWS? Unfortunately I've been receiving some PRs that haven't been tested at all. Most likely done with AI, trying to avoid asking that question though, but might have to start.

-Erik

[ekristen]

Due to your comment on the PR, is there a stopped JobStatus? If so we need to have a switch for it and error with so it gets filtered out as well.

[chrisns]

Thanks for catching that Erik! You're right — there is a STOPPED status (and STOP_REQUESTED too). I've pushed a fix that explicitly handles all terminal statuses: STOPPED, STOP_REQUESTED, COMPLETED, COMPLETED_WITH_ERROR, and FAILED — each with a descriptive error message rather than relying on the catch-all default.

[chrisns]

At a glance things look great, have one question on the filter for translation job.

Also I have to ask, have you fully tested these resources against AWS? Unfortunately I've been receiving some PRs that haven't been tested at all. Most likely done with AI, trying to avoid asking that question though, but might have to start.

-Erik

totally valid question, yes I confess I absolutely have used AI, however has been tested with a real account.
be neat if there was some way to prove a PR works without you having to run up a massive aws bill creating/destroying stuff to prove it🤔

[ekristen]

I'm a proponent of AI, I have zero issues with it, just so long as things are being actually tested against an account and mock tests are set to match. The inverse has been happening, tests are written and mocked by AI then everything fails on a real account. Not you, but other PRs I've received.

I do have some OSS credits from AWS, so I can test without too much fear of a high bill, but I appreciate it.