fix TestWalWritesFinalization: wait for the block record before reading the WAL

[samliok]

TestWalWritesFinalization read the WAL without waiting for it to be written, so it could observe an empty WAL and fail with "[]" should have 2 item(s), but has 0.

Start schedules block building on buildBlockScheduler, a separate goroutine. That task calls BuildBlock, which pushes the block onto the test builder's built channel and returns, so GetBuiltBlock unblocks the test there. The task then still has to take the epoch lock and call proposeBlock -> storeProposal, which is where the block record actually reaches the WAL. GetBuiltBlock returning is not a happens-before edge for that append, so the test could race ahead and read an empty WAL.

Fixed by waiting on AssertWALSize(2) before the read, which is what the same test already does before its second read and what TestWalWritesBlockRecord does before its own.

The test's third read needs no wait: that finalization record is appended by persistFinalization on the synchronous HandleMessage path, so it is written before InjectTestFinalizeVote returns.

Pre-existing on main, not introduced by any open PR. Surfaced in CI on an unrelated PR's run.

Verification

Reproduced on clean origin/main by wrapping the test body in a 20-iteration loop and running under -race with GOMAXPROCS=4; it failed on iteration 6 with the same message. With the fix, 3 runs x 20 iterations were clean, then the loop was removed and the test re-run clean.