fix(autoscaling): full Update field set, Describe filters + pagination, dup-name guard

[vieiralucas]

Summary

Cycle-4 bug-hunt Tier-1 cluster for the EC2 Auto Scaling service (findings 1.1, 1.9, 1.15, 4.5).

• UpdateAutoScalingGroup now persists every mutable field DescribeAutoScalingGroups echoes back (DefaultCooldown, HealthCheckGracePeriod, VPCZoneIdentifier, AvailabilityZones, NewInstancesProtectedFromScaleIn, ServiceLinkedRoleARN, TargetGroupARNs, LaunchTemplate↔LaunchConfiguration swap). Previously only Min/Max/Desired/LC/HealthCheckType applied → Terraform changes to anything else returned 200 then silently dropped → perpetual diff / "provider produced inconsistent result after apply".
• DescribeAutoScalingGroups / DescribeTags honor Filters (tag:/tag-key/tag-value/auto-scaling-group, and key/value/propagate-at-launch for tags). DescribeTags previously returned every group's tags → tag bleed across aws_autoscaling_group_tag resources.
• All Describe* paginate: MaxRecords clamped to the AWS cap, NextToken emitted on truncation and resumed.
• CreateAutoScalingGroup rejects a duplicate name with AlreadyExists instead of overwriting (which orphaned the prior group's launched instances).

Test plan

• 5 new unit tests (full-field update round-trip, dup-name AlreadyExists, tag filter, DescribeTags scoping, pagination). cargo nextest run -p fakecloud-autoscaling 7 pass.
• cargo clippy -p fakecloud-autoscaling --all-targets -- -D warnings clean.
• CI tfacc autoscaling shard exercises the read shapes against the real Terraform provider.

Surface

Behavior fix within an existing service; no new public API/SDK surface.

---

Summary by cubic

Fixes ASG drift and read inconsistencies: updates now persist all mutable fields, Describe APIs support filters and pagination, and duplicate names are rejected. This removes Terraform perpetual diffs and prevents tag bleed.

• Bug Fixes
  • UpdateAutoScalingGroup now saves all mutable fields (cooldowns, grace period, VPCZoneIdentifier, AZs, scale-in protection, ServiceLinkedRoleARN, TargetGroupARNs, and LaunchTemplate/LaunchConfiguration swap).
  • DescribeAutoScalingGroups and DescribeTags honor filters (tag:/tag-key/tag-value/auto-scaling-group; key/value/propagate-at-launch for tags) to stop tag bleed.
  • All Describe* calls paginate: MaxRecords clamped to AWS limits, NextToken returned and accepted.
  • CreateAutoScalingGroup rejects duplicate names with AlreadyExists instead of overwriting.

^{Written for commit e5741b0. Summary will update on new commits.}