Summary

This PR fixes several issues with classic (v1) ELB load balancers so they work the same way as v2 (ALB/NLB) load balancers.

What was wrong

• Missing fields on classic LBs: Querying availabilityZones, securityGroups, hostedZoneId, or vpc on a classic load balancer would fail with a type conversion error because those fields were never set during creation.
• Duplicate results: Every load balancer (classic and v2) appeared twice in query results due to being appended to the list before and after the tag filter check.
• Crashes with tag filters: When using --filters "tag:...", looking up security groups or VPCs could fail if those resources didn't match the tag filter, breaking the entire query.
• Unnecessary API calls: The targetGroups field on classic LBs tried to call the v2 API, which doesn't recognize classic ARNs. Now it returns an empty list immediately.
• Init function called for wrong resources: The ELB init function was being triggered for non-ELB resources (like S3 buckets or IAM users), wasting API calls. Now it exits early if the ARN isn't an ELB.
• Discovery region: The discovery code was parsing the region from the ARN instead of using the field already on the resource.

Changes

• aws_elb.go: Added missing fields to classic LBs, removed duplicate appends, made SG/VPC resolution graceful, added early exit for non-ELB ARNs in init, added classic LB guard on targetGroups.
• discovery.go: Simplified region handling to use f.Region.Data directly.