Keep eth.header_cids as a standard table, not hypertable.

[telackey]

Anytime we need to lookup by block hash, we join on the eth.header_cids table. However, if the table is partitioned by block_number, we make looking up by hash less efficient, since we also partition the index and need to check each one.

As this is by far the smallest major table (1 row per block, averaging about 1KB per row), the need for partitioning is already doubtful. Combined with slower lookups by hash, the performance impact tips negative rather than positive.

[roysc]

Nice, then we can also drop the header_result type; actually the get_child function can also go, since it was only used by canonical_header_from_array.

[roysc]

LGTM.

NB. this and #138 will functionally conflict (but git may not detect it).

[i-norden]

This is great, thanks! Is this waiting on additional changes?