Fix cluster node selection for SCAN, SSCAN, HSCAN, ZSCAN

[cmcinnes-mdsol]

We found that while using SSCAN on a stable cluster (where no hash slots are being moved from one node to another) we were getting a lot of MOVED responses.

Upon investigation we saw that for SSCAN the client was selecting a random writer node. I may be missing the original intent of the removed branch of the case statement but it seems like we shouldn't care about the position of the match option as the argument for it is a pattern not a key. Before this change Cluster#find_node_key was returning nil for all flavors of scan which was causing Cluster#find_node to select a random node. Tested out this branch and the MOVED responses went away while SSCAN continued to function normally.

[cmcinnes-mdsol]

thanks for the quick merge! do we think this is worth a changelog entry? if so can open another pr