fix: flaky test for gh_6634

Current test seems to be flaky. Previously it checked that we print
correctly in the logs the address of the replaced tuple during GC.
For some reason the test was searching for the first string with
`tuple_delete` in the log and expected it to be the one we are
looking for.

Current fix makes it search for exactly the deleted address we need,
not the first one appear in the log.

NO_DOC=test fix
NO_CHANGELOG=test fix