Block store duplicates lots of data

[AndrewSisley]

There is a lot of duplicate data held in the blockstore, and we can considerably shrink the size if we want to.

Note: The block store accounts for the vast majority of Defra's storage requirements. Bigger blocks also means slower writes, reads, and more network traffic.

These issues also permit structural data-divergance, where the duplication of data creates a structure where the data on the parent may diverge with the data on the child. It also prevents parents from containing more than one unique value (e.g. a composite of multiple fields, or multiple documents).

Testing also finds that property name length does impact the storage size given our current encoding, we may wish to reduce this too (or, less likely, change the encoding so field name has no impact on size).

Tasks

Preview Give feedback

1. Remove Name from block links #3056
   area/db-system code quality perf refactor +4
2. SchemaVersionID is held on composite and field level blocks #3086
   area/db-system perf +2
3. DocID is held on composite and field level blocks #3087
   area/db-system perf +2
4. FieldName is held on composite and field level blocks #3089
   area/db-system perf +2