Remove the --unsafe-pruning CLI-argument (step 1) (#10995)

* sc-client-db: utils::open_database(...) — return OpenDbError so that the caller could tell the `OpenDbError::DoesNotExist` clearly

* sc-client-db: utils::open_database(..) — accept the `create: bool` argument

* sc-client-db: pruning — optional argument in the DatabaseSettings

* sc-state-db: Split `Error<E>` into separate `Error<E>` and `StateDbError`

* StateDb::open: choose the pruning-mode depending on the requested and stored values

* sc-state-db: test for different combinations of stored and requested pruning-modes

* CLI-argument: mark the unsafe-pruning as deprecated

* Fix tests

* tests: do not specify --pruning when running the substrate over the existing storage

* fix types for benches

* cargo fmt

* Check whether the pruning-mode and sync-mode are compatible

* cargo fmt

* parity-db: 0.3.11 -> 0.3.12

* sc-state-db: MetaDb::set_meta — a better doc-test

* cargo fmt

* make MetaDb read-only again!

* Remove the stray newline (and run the CI once again please)

* Last nitpicks

* A more comprehensive error message

substrate/client/api/src/client.rs

@@ -146,6 +146,9 @@ pub trait BlockBackend<Block: BlockT> {
 	fn has_indexed_transaction(&self, hash: &Block::Hash) -> sp_blockchain::Result<bool> {
 		Ok(self.indexed_transaction(hash)?.is_some())
 	}
+
+	/// Tells whether the current client configuration requires full-sync mode.
+	fn requires_full_sync(&self) -> bool;
 }
 
 /// Provide a list of potential uncle headers for a given block.