Replace cl_audits table with tree_transactions

[kespinola]

Context

A cl_audit system was introduced in this PR #94 which keeps complete revision history of change logs for a tree where the cl_items tracks the current change history for the tree. The cl_audit records also record the corresponding transaction signature and instruction identifier.

The cl_audit records in upstream is a configurable feature that only adds the records but Helius has used these records to provide a getSignaturesForAsset method which returns a list of transaction signatures related to a leaf. https://github.com/rpcpool/digital-asset-rpc-infrastructure/blob/triton-build/digital_asset_types/src/dapi/signatures_for_asset.rs#L21

Issues

• The cl_audits significantly increase the storage requirements for the indexer because its a complete revision history of change log items.
• There is a bug causing duplicate cl_audit records [Bug] saving cl_audit records is not idempotent  #152

Proposal

Replace the writing of cl_audit with tree_sequence records. The tree_sequence tracks every sequence for a tree seen by the indexer with additional meta information like the transaction signature and instruction identifier (as currently recorded on cl_audits). Instead of writing n change log records per transaction with each sequence change write 1 record for the sequence change only.

CREATE TABLE tree_sequences (
    -- Composite primary key consisting of the sequence number and tree public key
    seq BIGINT NOT NULL,
    tree BYTEA NOT NULL,
    -- Index of the leaf causing the tree change
    leaf_idx BIGINT,
    -- Signature of the transaction associated with the instruction
    signature BYTEA NOT NULL,
    -- Identifier for the instruction
    instruction VARCHAR,
    PRIMARY KEY (seq, tree)
);

Uses

• Maintain feature parity with the existing getSignaturesForAsset RPC method by providing transaction signatures for tree leaves.
• Enhance tree backfilling and reporting capabilities by utilizing sequence information
  [Feature] Tree Sequence Based Backfilling #150

[NicolasPennie]

Discussed reconciling this schema with Helius' new one here: #150 (comment)

[kespinola]

resolved by #163