fix: start the stack from a fresh clone

[YvesLepidus]

On a fresh clone of main, make up fails for the site and Nextcloud stacks, and the WordPress → Nextcloud integration cannot reach Nextcloud.

Changes

• Site: SITE_COMPOSE now uses env UID=... docker compose .... In bash, which is /bin/sh on Fedora, UID is readonly, so UID=... docker compose failed with UID: readonly variable.
• Nextcloud services: mysql was renamed to database in _refresh-nextcloud-images and _start-nextcloud. The nextcloud-development bump in 7f556a7 renamed the service, which made the Makefile fail with no such service: mysql.
• Nextcloud port binding: IP_BIND=0.0.0.0 was added to NEXTCLOUD_COMPOSE. The same bump made ports bind to 127.0.0.1 by default, so WordPress could no longer reach Nextcloud through host.docker.internal:8082. This restores the previous behavior.
• Trusted domain: a new _set-trusted-domains step adds host.docker.internal to Nextcloud's trusted_domains. Without it, requests from WordPress got Trusted domain error.

Validation

Tested with make up on a fresh clone (Fedora 44, bash 5.3):

• WordPress (:80), the site (:8081) and Nextcloud (:8082) all respond 200.
• wordpress_login_backend, admin_group_manager and groupquota are enabled, and the admlibrecode group is created.
• curl http://host.docker.internal:8082/status.php from the WordPress container returns 200.

Notes

• With IP_BIND=0.0.0.0, the Nextcloud database port (3307) is also published on all interfaces.
• Not fixed here: woocommerce-nextcloud-admin-group-manager requires WordPress 7.0, but wordpress-docker:latest ships 6.9.4, so the plugin is not activated. make up ignores this error.

[YvesCesar]

Superseded by #24.