feat: honour Retry-After for artifact retries

[lox]

Summary

This follow-up teaches artifact retry loops to honour server-provided Retry-After headers, so retry pacing can be coordinated by the backend during pressure events.

What Changed

1. Added a small shared helper in internal/artifact/retry_after.go to parse and apply Retry-After values to retry intervals.
2. Wired Retry-After handling into CreateArtifacts retries in internal/artifact/batch_creator.go.
3. Wired Retry-After handling into UpdateArtifacts retries in internal/artifact/uploader.go.
4. Added tests in internal/artifact/batching_test.go for valid and invalid Retry-After behaviour.

Why

1. Improves stability by letting the server shape retry timing during load.
2. Reduces risk of synchronized retry bursts from many agents.
3. Keeps existing retry behaviour intact when Retry-After is absent or invalid.

Validation

1. go tool gofumpt -extra -w internal/artifact/retry_after.go internal/artifact/batch_creator.go internal/artifact/uploader.go internal/artifact/batching_test.go
2. go test ./internal/artifact -count=1
3. go test ./api ./agent ./clicommand -count=1

Notes

1. This PR is stacked on top of feat: support server-hinted artifact batch sizing #3736.

[moskyb]

looks good! good to bring this in line with many of our other retry loops.