ci: run the integration tests against Redis 7, not Redis 3

[jqnatividad]

Since the 2026-10-06 nightly, every integration run (on main and on PRs) has failed at Initialize CKAN database, before DP+ runs:

File ".../redis/connection.py", line 1186, in on_connect_check_health
    self.handshake_metadata = self.read_response()
redis.exceptions.ResponseError: unknown command 'HELLO'

Cause: pydocket, which Prefect depends on, now requires redis>=6. pip therefore upgrades ckan-dev's redis-py from 5.0.7 (last green run, 2026-10-05) to 8.1.0 (first red run, 2026-10-06, same main commit). redis-py 8 sends HELLO when it connects, and HELLO arrived in Redis 6.0. The CI service was redis:3.

Fix: use redis:7 in ci.yml and main.yml, which docker-compose.integration.yaml already uses.

I reproduced it locally: redis-py 8.1.0 → redis:3 gives ResponseError: unknown command 'HELLO', and redis:7 gives PING → True. This PR's own integration run is the end-to-end check.

Deployment note: this isn't CI-only. A fresh DP+ install now resolves redis-py 8.x (through Prefect → pydocket), and that won't connect to a Redis server older than 6.0.

🤖 Generated with Claude Code