Store state trie in SQLite

This means that writes of blocks and writes to state can be kept
in sync, because the write barriers that SQLite provides will
apply to both forms of data. This should eliminate
inconsistencies between the two data stores.

Note that I also renamed the SQL database from `blockdata.db` to
`db.sqlite3`.

Author: JamesHinshelwood

eth-trie.rs/src/db.rs

@@ -36,9 +36,6 @@ pub trait DB: Send + Sync {
         Ok(())
     }
 
-    /// Flush data to the DB from the cache.
-    fn flush(&self) -> Result<(), Self::Error>;
-
     #[cfg(test)]
     fn len(&self) -> Result<usize, Self::Error>;
     #[cfg(test)]
@@ -84,10 +81,6 @@ impl DB for MemoryDB {
         Ok(())
     }
 
-    fn flush(&self) -> Result<(), Self::Error> {
-        Ok(())
-    }
-
     #[cfg(test)]
     fn len(&self) -> Result<usize, Self::Error> {
         Ok(self.storage.try_read().unwrap().len())

z2/src/zq1/db.rs

@@ -107,10 +107,6 @@ impl DB for StateDatabase {
 
         Ok(())
     }
-
-    fn flush(&self) -> Result<(), Self::Error> {
-        Ok(())
-    }
 }
 
 pub struct Db {