• Introduces a dedicated url field on ContentNode and a persistent UrlIndex, so pages can be looked up by their public URL instead of resolving the file path first (ContentResolver, ViewResolver, NavigationFunction now use
  Content#byUrl). Content#byUri is deprecated in favor of byPath/byUrl.
• Adds a persistent SectionIndex so listSectionEntries no longer scans the whole metadata tree, and section lookups now work correctly even when a page has a custom url.
• Makes metadata updates incremental: file system events are now debounced and batched through a new ContentChangeCoordinator, and single file/directory changes update only the affected metadata instead of triggering a full resync. Directory
  renames are handled as a delete+create of the affected subtree. A watcher OVERFLOW event now triggers a full resync instead of being silently dropped.
• Removes the in-memory metadata backend (MemoryMetaData/MemoryQuery) and the queryIndexMode switch; PersistentMetaData is now the only backend.
• Makes YAML header writes and the CLI pid file write atomic (write-to-temp + move), and the pid file now also stores the process start time so a reused PID isn't mistaken for a still-running CMS process.
• Bumps CI to JDK 25 (release) instead of 25-ea.

Fix included in this PR

• Fixed a bug in ContentResolver#getContent where a content node with its own redirect.location/redirect.status set was returned as a normal (empty) DefaultContentResponse instead of a RedirectContentResponse, because an earlier
  isRedirect() check already returned before the actual redirect-response branch was ever reached. The redirect and alias-redirect handling has been merged into a single if/else if chain so the correct branch is no longer unreachable.
• Removed now-dead code: the unreachable comment block in ContentResolver, and the unused getUrl(ReadOnlyFile) helper (and its now-unused imports) in AbstractCurrentNodeFunction, which was made obsolete by the switch to
  ContentNode#url().