Change bytecode storage to be by hash as opposed to by address

[non-fungible-nelson]

Today, code storage is stored by account address, but we need to store by code hash. Duplicated smart contracts can point to the same hashed bytecode. We also must serve data by hash in the context of the snapsync server.

• Do a comparison with a full sync node to compare the improvement in storage reduction
• Measure the performance impact of the change on syncing and block import

• Change code storage to be stored by code hash instead of account address
• Backwards capability for code storage by account
• Compare stored code values from existing column family to new column to ensure consistency
• Performance test syncing to ensure read/write performance caused by changes to tracking code hashes by address have minimal impact
• Experimental flag to enable code storage by code hash
• Compare storage size taken by new column families with old storage mode
• Track code hashes by address so that self-destruct can removed when all code storage is unreferenced

[jframe]

Mainnet code by address hash column size

Column Family	Keys	Column Size	SST Files Size
CODE_STORAGE	37701832	10 GiB	12 GiB

Mainnet code by code hash column size

Column Family	Keys	Column Size	SST Files Size
CODE_STORAGE	1231697	3 GiB	3 GiB

We save ~7GB by storing by code hash