Implement the migrate iterator for iterating key values

[git-hulk]

Currently, we need to iterate all keys in the database in different places like the cluster migration and kvrocks2redis, but don't have an iterator for this purpose. It's very error-prone to implement this in different places since Kvrocks may add a new column family in the future, and we must be careful to iterate all keys in all column families. This would be a burden for maintenance, So I think we could add an iterator to iterate all keys in all column families.

Proposal

For the API, we could generally comply with the rocksdb's iterator API, but we allow to add some extra functions if needed.

class Iterator {
public:
    Iterator(Storage *storage, const rocksdb::ReadOptions &options, const int slot = -1);
    ~KeyIterator();

    bool Valid() const;
    void Next();

    rocksdb::WriteBatch *Batch() const;
    Slice Key() const;
    Slice Value() const;
    RedisType Type() const;
};

And when implementing this iterator, it will iterate the metadata column family first and check its type, if it's not a string, then it will iterate the corresponding column family to get subkeys. That said, if we have a key foo with type hash, then the iterator will iterate foo and foo:field1, foo:field2, and so on.

This solution can bring those benefits:

• The codes look more intutive
• Can reuse this iterator if we want to iterate keys only

[git-hulk]

@caipengbo Would you mind taking a look while you're free.

[caipengbo]

Would you mind taking a look while you're free.

Good job, LGTM!