Notes from reveiw of 2.1 encodings

The following are notes I took while reviewing 2.1 encodings in preparation for making 2.1 stable.

• The binary mini block encoding does not compress offsets and it should
  • High priority: offsets are likely very compressible (delta and bit packing in particular)
• The binary mini block encoding should allow general compression to be applied to the values buffer
  • Low priority: mini-block level compression is probably fine
• Mini block inline bitpacking should allow the inner compression to be specified (e.g. bitpacking followed by general compression should be possible)
  • Low priority: mini-block level compression is probably fine
• The general-per-value encoding should support setting the compression level
  • High priority: this is an interesting config that should be exposed
• The general-per-value encoding should defer to the binary encoding
  • Medium priority: If we don't do this we won't have compression of the offsets, but offsets compression isn't terribly useful in per-value situations
• Byte stream split could support underlying compression
  • Low priority: mini-block level compression is probably fine
• We are not compressing rep/def in miniblock
  • High priority: rep/def is quite large if not compressed at all, this is wasteful

Start by locating the implementations for binary mini block, general-per-value, byte stream split, and repetition/definition-level encodings. Review how offsets, values, compression configuration, and rep/def data are handled, then prioritize the high-priority items in the note. Done means the selected compression behaviors are implemented and verified by tests.