[3.1.2 backport] Matching databases for different config groups not being assigned same metadataID

When upgrading to 3.1.x, the same database under different config groups can end up with different metadataIDs.  This results in the database being forced offline (requires resync), but resync won't actually fix the situation (it will just force the database in the other config group offline).

Steps to repro:

• Bucket must have existing _sync:seq and at least one document

• start SG in config group 1
• create database 'foo' for bucket 'bar'

(metadata is set to _default)

• start SG in config group 2
• create database 'foo' for bucket 'bar'

(metadata is set to foo, requires resync)

The fix should ensure that the same database in different config groups ends up with the same metadataID, and should also evaluate repair mechanisms for existing cases that hit this.