Session commands with max_slave_connections=0 after switchover do not discard stale connections

Similar to MXS-5209, where reads did not discard stale connections, session commands and other "special" route targets did not cause stale connections to be discarded.

This happened because only reads got the special treatment when max_slave_connections=0 was used and session commands were explicitly excluded. Given that it doesn't really make sense to delay the discarding, the solution to this is to just treat everything as if it was a write which will cause stale connections to be discarded.