What this PR does / why we need it?

This PR contains two part:

1. Make the storage and executor backend read-write separation.
2. Merge the four RocksDB instance into one

The part 1 is the precondition of part 2.

Read-write separation

1. Divide Storage trait into ReadOnlyStorage and Storage. The ReadOnlyStorage contains all the get* methods and the Storage trait inherit the ReadOnlyStorage trait.
2. Divide ExecutorAdaptor trait into ReadOnlyExecutorAdaptor and ExecutorAdaptor. The difference is that the ExecutorAdaptor requires the ApplyBackend trait.
3. The consensus module and executor module use Storage and API module and mempool module use ReadOnlyStorage
4. The consensus module use ExecutorAdaptor and API module, interoperation module and mempool module use ReadOnlyExecutorAdaptor

Merge RocksDB instance

This PR abstracts the adapter module in the core/storage crate into a new standalone crate named core/db. The new crate contains memory DB and RocksDB wrapper functionalities. Both core/storage and core/executor now directly depend on core/db, which is a change from the previous setup. The former state_db, metadata_db, and header_cell_db are now represented as c8, c9, and c10 column families within the main RocksDB instance. This restructuring enhances modularity and efficiency in managing database components.

What is the impact of this PR?

⚠️This is a breaking change and have to clear the data.

The previous data folder layout:

├── axon.ver
├── consensus_wal
├── peer_store
│   ├── addr_manager.db
│   ├── ban_list.db
│   └── tmp
├── rocksdb
│   ├── block_data
│   ├── state_data
│   └── system_contract
│       ├── header_cell
│       └── metadata
└── txs_wal

The current data folder layout:

├── axon.ver
├── consensus_wal
├── peer_store
│   ├── addr_manager.db
│   ├── ban_list.db
│   └── tmp
├── rocksdb
└── txs_wal

Special notes for your reviewer

Because of this PR change so many files, review it commit by commit.

Which issue(s) this PR fixes: