fix(scale-up): return partial runner creation results

[edersonbrilhante]

Description

Refactor runner creation to return structured partial-success results instead of throwing ScaleError.

• Return created instance IDs with retryableErrorCount and nonRetryableErrorCount.
• Preserve successfully created EC2 instances when Fleet or dedicated-host requests only partially succeed.
• Classify configured scale errors, transient AWS errors, network failures, throttling, and server failures as retryable.
• Treat unclassified AWS errors as non-retryable.
• Propagate the result contract through EC2, pool, and scale-up providers.
• Convert unexpected provider exceptions into retryable result counts.
• Treat GitHub runner registration failures as retryable and terminate the affected EC2 instances.
• Retry only the corresponding number of SQS messages.
• Remove the obsolete ScaleError implementation and handler logic.
• Skip unsupported event types without retrying them.

Test Plan

TBD

Related Issues

None.

[github-actions]

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

Scanned Files

None

[npalm]

would rename here the const name, since the return value is now the result. So a more logical name woudl be createRunnerResult

[edersonbrilhante]

scaleUp still returning the same rejectedMessageIds. That is a list of messageId that will send to ReportBatchItemFailures.