Update metadata cache in sys.tables immediately upon relevant change

Right now the system tables have a lot of flags the `sys.segments` tables has a `is_published` and `is_available` etc. These are updated on different cadences due to caching that make it confusing the the user. It also makes a poor experience in the web console (which is a system table user). It would be really cool have a write through cache or something fancy like that so that the systems tables would both be fast and able to reflect the current up-to-date status of the cluster.

Start by tracing how the sys.segments system-table flags, including is_published and is_available, are cached and updated. Define the required consistency behavior for relevant cluster changes and verify that the system tables remain fast while reflecting current status for users and the web console.

Written by the indexing model from the issue text.