feat: Add Terraform destroy step to GitHub Actions workflow and update README for redeploy behavior

Author: 02loveslollipop

.github/workflows/deploy.yaml

@@ -138,6 +138,16 @@ jobs:
         run: |
           terraform init -input=false
 
+      - name: Terraform destroy existing instance
+        if: ${{ matrix.deployer == 'terraform' }}
+        working-directory: ${{ matrix.path }}
+        shell: bash
+        env:
+          TF_IN_AUTOMATION: 1
+        run: |
+          terraform destroy -input=false -auto-approve -target=aws_instance.mediamtx || true
+          terraform destroy -input=false -auto-approve -target=aws_eip_association.mediamtx || true
+
       - name: Terraform apply
         if: ${{ matrix.deployer == 'terraform' }}
         working-directory: ${{ matrix.path }}

infra/terraform/media_relay/README.md

@@ -11,6 +11,7 @@ This module provisions everything required to host the relay on a single EC2 ins
 - Optional Caddy reverse proxy that fetches Let's Encrypt certificates for the supplied domain and fronts the Mediamtx HTTP/WebRTC endpoints.
 - Elastic IP associated with the instance for a stable ingress point.
 	- If you supply `existing_eip_allocation_id`, the module reuses that Elastic IP instead of allocating a new one.
+	- The GitHub Actions workflow destroys the EC2 instance (and its EIP association) prior to each redeploy. Without an `existing_eip_allocation_id`, a fresh Elastic IP is allocated on every run.
 
 You can optionally set `vpc_id`, `subnet_id`, or `key_name` to place the instance in a specific network or enable SSH access.