Implement WAL iterator for migrating by WAL logs

[git-hulk]

No description provided.

[caipengbo]

In #2004, we implement the unified key-value iterators, I want to change data to provide a unified iterators.

For incremental data, we need to call rocksdb::WALIter and then implement the rocksdb::WriteBatch::Handler interface.

I want to provide an interface that behaves similarly to engine::DBIterator:

kvrocks/src/storage/iterator.h

Lines 52 to 81 in c772109

	class DBIterator {
	public:
	explicit DBIterator(Storage *storage, rocksdb::ReadOptions read_options, int slot = -1);
	~DBIterator() = default;
	bool Valid() const;
	void Seek(const std::string &target = "");
	void Next();
	// return the raw key in rocksdb
	Slice Key() const;
	// return the namespace and user key without prefix
	std::tuple<Slice, Slice> UserKey() const;
	Slice Value() const;
	RedisType Type() const;
	void Reset();
	std::unique_ptr<SubKeyIterator> GetSubKeyIterator() const;
	private:
	void nextUntilValid();
	Storage *storage_;
	rocksdb::ReadOptions read_options_;
	int slot_ = -1;
	Metadata metadata_ = Metadata(kRedisNone, false);
	rocksdb::ColumnFamilyHandle *metadata_cf_handle_ = nullptr;
	std::unique_ptr<rocksdb::Iterator> metadata_iter_;
	std::unique_ptr<SubKeyIterator> subkey_iter_;
	};

.

It will wrap rocksdb::WALIter and can return different value types. It is possible to implement their processing logic depending on the concrete type.

HDYT @git-hulk

[git-hulk]

@caipengbo Sounds good to me.