Fix high data consumption issue

[rounakdatta]

The root cause of the very high data consumption (about ~5GB per hour) is as follows:

• The Immich machine learning container downloads a face recognition model of about 600MBs.
• Now, although it wouldn't try to fit the entire model into memory, but still it derives inference using the model.
• However, the container has been assigned only the default of 300MB of memory. And hence if there is any load at all, it would crash.
• The container restarting means it would download the same model again.
• This happening multiple times in an hour led to the very high data wastage.

To mitigate, we have temporarily given the machine learning model enough memory (~2GB dedicated for now), which we'll re-evaluate later. Also, we have disabled MongoDB, Cronicle and NTFY, given we are no longer using these services actively.