Simplify GCP prod Terraform config for minimal, non-HA instances

[WiedersehenM]

Summary

This PR simplifies the GCP prod Terraform configuration and adjusts defaults to use the smallest reasonable instance sizes with no HA / replicas, aligning with the course requirement for a cost‑efficient, easy‑to‑understand setup.

Key work is focused under deployment/envs/gcp/prod (and related modules).

---

What’s Changed

1. Prod terraform.tfvars and example

• Reduced terraform.tfvars to only the values that must differ per environment:
  • project_id, base_domain
  • GCS bucket name
  • Cloud Run service names, images, and basic env vars
  • Cloud SQL instance/database name + credentials
  • Redis instance name
  • Filestore instance name (optional)
  • Rails master key
• Updated terraform.tfvars.example to mirror the same structure as prod tfvars but with placeholder values.
• Documented testing option 1 for Cloud SQL:
  • cloud_sql_enable_private_service_connection = true
  • cloud_sql_enable_public_ip = true
  • This enables public IP access while keeping private networking, avoiding empty network references.
• Added vpc_connector_cidr to example (e.g. 10.8.0.0/28) so that the Serverless VPC connector can be created without 400 errors.
• Disabled Filestore by default in example:
  • enable_filestore = false
  • Rationale: Filestore is optional for this project and often blocked by additional network/permission requirements; defaulting to off makes the example easier to apply.

2. Instance sizing and HA/replicas

• Cloud SQL
  • Default tier set to the smallest tier used in this project (e.g. db-f1-micro).
  • Reduced default disk size to 10GB.
  • Availability type set to ZONAL (non‑HA).
  • Backups disabled by default for teaching / minimal‑cost scenarios.
• Redis (Memorystore)
  • Tier set to BASIC instead of STANDARD_HA.
  • Default memory size reduced to 1GB.
• Cloud Run
  • Keep small resource limits (1 CPU / 512Mi) and rely on module defaults for min/max instances.
  • Removed redundant overrides from tfvars (no extra replicas configured explicitly).

3. Networking and VPC Connector

• Kept private networking (enable_private_networking = true) for Cloud SQL and Redis by default.
• Added explicit vpc_connector_cidr to avoid “CIDR required” errors when creating the Serverless VPC Access Connector.
• Updated modules/gcp/vpc_connector:
  • Set min_throughput = 200 and max_throughput = 300 to satisfy the API’s throughput constraints (multiples of 100 between 200 and 1000 Mbps).

---

Motivation

• Teaching / lab environment: we want students to be able to terraform apply with minimal friction.
• Cost efficiency: smallest possible instances, no HA, and no unnecessary replicas.
• Config clarity: move generic defaults into variables.tf and keep tfvars focused on per‑environment differences only.
• Resilience to common errors:
  • Avoid Filestore/network errors by disabling Filestore by default.
  • Avoid VPC Connector 400s by providing a valid CIDR.
  • Provide a documented pattern for enabling Cloud SQL public IP while retaining private networking.