don't use Flag API to load flagging ID in test

This came up when I was running tests with older versions of the code, to get a better understanding of what was changed by #2154251: Flag status incorrect when flagging another flag as a reaction to unflagging (Javascript).
We should be using a direct DB query to get a flagging ID, not relying on the API, as that's what we're meant to be testing.

Will commit a fix.