fix(horde): load OAuth admin page when OAuth is not configured

[TDannhauer]

Summary

• Make RouteUrlWriter optional in OAuthProviderController so list/edit pages can load without full URL-generation dependencies
• Show the existing storage-unavailable notice when the null OAuth provider repository is active
• Fall back OAuth callback URLs to the webroot path when named routes are unavailable

Motivation

The OAuth provider admin page should be usable on instances without OAuth configured. Controller construction and the list template should degrade gracefully instead of failing hard when storage or route URL generation is incomplete.

This complements horde/Core#198 (DbServiceFactory Unix-socket fix), which resolves the fatal PDO error on PostgreSQL socket deployments.

Changes

• OAuthProviderController: optional RouteUrlWriter, resolveCallbackUrl(), storage-unavailable view data
• templates/admin/oauthprovider/list.html.php: notice when provider storage is unavailable

Test plan

• OAuth provider list renders HTTP 200 with null repository forced
• Storage-unavailable notice appears when NullOAuthProviderConfigRepository is active
• Manual: open /horde/admin/authentication/provider/ without OAuth providers configured

[ralflang]

I am not really convinced - Nothing in RouteUrlWriter depends on Oauth being configured so it's questionable whatever this fixes for you is really the described behaviour.

Please document how exactly RouteUrlWrite becomes unavailable by disabling / not configuring OAuth.

Generally:

The UrlWriter subsystem is fundamental - if it is not available, we should bail out in a "system temporarily not available" fashion regardless of which endpoint was asked for rather than try to rescue this.

Also if DB is configured but unavailable this is fundamental and "gracefully" starting the system with some resources available and others seemingly disappeared (rather than "currently not available" if the protocol supports it) is actively dangerous in anything synchronizing/consuming

I agree the oauth admin page should survive conditions related to OAuth not yet being configured sufficiently.

[TDannhauer]

Closing after re-investigation — the review was right on both counts.

1. RouteUrlWriter is never unavailable in real web requests. The failure I attributed to it came from a CLI test harness running without HTTP_HOST; in the actual request path RouteUrlWriterFactory always resolves. Nothing here depends on OAuth being configured.
2. The real fatal was the Unix-socket config bug fixed in fix(core): preserve Unix socket SQL config in DbServiceFactory Core#198 (PermissionService -> DbServiceFactory forcing localhost:3306 onto a PostgreSQL socket deployment). With that fix, this page loads fine on an instance without OAuth configured: horde_oauth_providers is simply empty and the list renders with no providers — no controller changes needed.
3. A deployment whose horde_oauth_providers migration never ran should fail loudly, consistent with every other admin page (perms, prefs, ... would be equally broken there).