Drop points and postings in favor of skiplists for dimension fields

This is expected to help with storage efficiency and indexing throughput, but may penalize query performance. The challenge is that dimension values are not sorted but clustered, as they're used to calculate _tsid that is the primary sort field for TSDB.