Add direct EC2 orchestration for distributed Presto benchmarks

[GregoryKimball]

Summary

• add an SSM-managed EC2 lifecycle for dedicated coordinators and distributed native workers
• support cache-off cold suites, AsyncDataCache warm suites, telemetry, artifact collection, and controlled CPU tuning
• support role-specific AMIs and images for mixed x86/ARM fleets

Depends on #376. Until that PR merges, this branch also contains its S3 commits; the final diff should be reviewed after rebasing onto main.

Test plan

• python3 -m unittest presto.aws.ec2.test_aws_cluster
• bash -n presto/aws/ec2/remote/configure_and_start.sh presto/aws/ec2/remote/collect_node.sh
• 8/16/32-worker M7a and M9gd cold SF1000 qualification
• 8-worker M9gd cache-off, memory-cache, and memory+NVMe qualification
• 16/32-worker M9gd memory-cache qualification
• G7e GPU worker qualification at 1/2/3/4 workers

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

[GregoryKimball]

Closing at the author's request; development will continue on a fork branch before a PR is opened.