Consider removing schema hash from the header

[agavra]

Currently the Imprint data format has a "schema" field which is 8 bytes: 4 bytes for the fieldspace and 4 for the schema hash. The idea behind this was that the schema hash would allow for more efficient validation of operations.

For example, consider three records with the following schema:

1. Fields 1,2,5,6
2. Fields 1,2,8,9
3. Fields 2,4,7

A function that requires Fields 1,2 would accept records 1 and 2 but not 3. Maintaining hashes let us build up a prefix-trie of schema hashes (the leaf nodes are hashes, the internal nodes are field ids) that can efficiently determine superset/subset/disjoint conditions when comparing two schemas without scanning the entire field directory for each.

It also allows checking whether two records are of the same schema in O(1).

The prefix-trie lookup is ~O(log N) where N is the total number of fields while a direct field dict comparison is O(M+L) where M is the number of fields in the first record and L is the number of fields in the second.

All things being equal, it seems like maintaining a schema hash is not so useful for common operations and just adds unnecessary complexity.