Record clustered index in index usage

[YangKeao]

Enhancement

Now, only a nonclustered primary key is recorded in the index usage. For a clustered index (which is usually read through TableReaderExecutor), it's not recorded. It'd also helpful to include the statistics for the clustered index in the index usage, because it can help the user to decide whether the current nonclustered primary key is good (though altering it is hard).