Fixes the probe arity in sabotage notes

[johnnyt]

Two sabotage notes in test/encryptor/ecto/migrator_run_test.exs named probe/2: the note above "a second run finds every row already in the target state" and the note above "a second adoption pass finds the row already backfilled". The function both mutations target is the private probe/3 in Encryptor.Ecto.Migrator.Pass (its @spec probe(t(), MapSet.t(identity()), binary() | nil)); both notes now say probe/3. load_probe/2 is a different function and is untouched.

Test comments only; no library change and no changelog fragment (the fragment README excludes test harness changes).

Checks

• grep -rn "probe/2" test/ returns nothing on this branch; before the change it returned the two notes above.
• load_probe/2 appears in lib/ and docs/ the same number of times as on main, and never in test/.
• Full mix quality green on the committed tree, database arm included: 891 of 891 tests passed, Credo, docs and Dialyzer clean. The committed tree is byte-identical to the tree the gate ran on.

Review

In-turn review, gate tier: the diff is two comment lines in one test file and touches no contract surface. I re-read both notes against Pass: the first mutation (answer :not_target unconditionally) is a change to probe/3's clauses, and the second (read the source column instead of the target) is a change to the value probe/3 is handed, so probe/3 is the right name in both. Every probe/ mention left in the moduledoc and ADR-0002 already carries arity 3 or names load_probe/2 correctly.