Is your feature request related to a problem? Please describe.
Sorting is very important for Parquet Read performance. This is because Parquet file has builtin min/max index which can be leveraged to filter out pages.
Today, Parquet Converter only has 1 primary sorting column which is metric name. For different usecases, if users know that there is another common column in their query such as cluster, namespace, etc. We should allow users to configure additional sorting columns other than metric name to provide better query performance.

Describe the solution you'd like
Add a new flag for additional sorting columns in https://github.com/cortexproject/cortex/blob/master/pkg/parquetconverter/converter.go#L142.

This can be a runtime config that can be overriden per tenant