fix: switch ECS services to Fargate Spot to reduce cost

[Akash29g]

What

Switch both ECS services (docanalytics-api-svc, docanalytics-web-svc) from on-demand Fargate to Fargate Spot capacity.

Why

Cost Explorer (Charge type = Usage, gross pre-credit) shows Elastic Container Service as the single largest daily line item on the account, running ~$2/day combined with ALB and RDS. With AWS credits nearly exhausted, Fargate Spot gives up to ~70% off the same vCPU/memory pricing for both services, with no code or architecture change required.

Change

• infra/ecs.tf → aws_ecs_service.api and aws_ecs_service.web: replaced launch_type = "FARGATE" with a capacity_provider_strategy block (FARGATE_SPOT, weight 1, base 1).
• Added aws_ecs_cluster_capacity_providers.main so the cluster allows Spot capacity, with a matching default strategy.

Scope / Safety

• Infra-only change — no application code, Dockerfiles, or task definitions affected.
• desired_count = 1 and the lifecycle { ignore_changes = [task_definition, desired_count] } blocks are unchanged, so the existing CI/CD deploy flow (deploy.yml) is unaffected.
• Fargate Spot tasks can be reclaimed with a 2-minute warning; ECS reschedules automatically. At desired_count = 1 this may cause a brief (seconds-level) blip, not sustained downtime.
• deploy.yml does not run Terraform, so this change requires a manual terraform apply from infra/ after merge — it will not take effect on push alone.

How to verify

1. terraform plan locally shows changes limited to the 2 services + 1 new capacity-provider resource — no destroys.
2. After terraform apply: ECS console → Services → both services show CAPACITY PROVIDER: FARGATE_SPOT and RUNNING count stays at 1 throughout the rollout.
3. docanalytics.dev stays reachable during and after the rollout.
4. Next day: Cost Explorer (Charge type = Usage, grouped by Service, Daily) shows a drop in the Elastic Container Service line vs. the prior day's baseline.