Add queue-based incremental STRM refresh

[hrd201]

Summary

• Add a queue-based incremental STRM refresher that keeps a local SQLite index and refreshes only candidate AList directories.
• Add missing .strm regeneration when Emby deletes a local STRM file but the remote media still exists.
• Add generic example config, task wrapper, detailed docs, and README/README.zh-CN/CHANGELOG updates.
• Ignore real incremental refresh configs and SQLite state files to avoid committing AList tokens or runtime state.

Privacy

• The committed config uses generic placeholders only (YOUR_ALIST_HOST, YOUR_ALIST_TOKEN, /mnt/cloud/...).
• No server hostname, IP address, AList token, or real media directory from the test deployment is included in the added files.

Validation

• python3 -m py_compile scripts/incremental_strm_refresh.py
• bash -n scripts/task_incremental_refresh.sh
• Loaded config/incremental-strm.json.example and verified path mapping to {media_path} output.
• Searched the worktree for deployment-specific host/IP/token markers before commit.

Supersedes #1 because that PR used a branch name with the codex/ prefix.

[hrd201]

Follow-up fix pushed in e822e19.

What changed:

• The incremental refresher now continues consuming newly queued child directories within the same max_dirs_per_run budget.
• Child directories are queued whenever a refreshed directory contains them, not only when they are first seen.
• The script no longer immediately requeues the directory currently being scanned after generating a .strm, avoiding duplicate same-run requests.

Server validation:

• Deployed the updated script to /opt/xstrm-suite/scripts/incremental_strm_refresh.py.
• Tested with parent directory /mnt/115/剧集/凡人修仙传 (2020) using a temporary output directory and temporary SQLite state.
• The run reached Season 2 from the parent directory and generated the expected temporary .strm for episode 181.
• incremental-strm-refresh.timer remains active and will use the updated script on the next scheduled run.