Guide users to consider adding an index on their TPH discriminator column

We don't automatically create an index on the discriminator, since that doesn't necessarily make sense in all cases (https://github.com/dotnet/efcore/issues/4030).

However, we should make users aware of this, and discuss when they might want to do this.

Note: for SQL Server specifically, it may make sense to make the index clustered. SQL Server prefers table scans to using non-clustered indexes which aren't highly selective, and discriminator value cardinality is generally low, so low selectivity (see https://stackoverflow.com/a/17861246/640325).