test: cover the write rollback's success path

[michael-denyer]

Closes the gap #127 surfaced by mutation testing: deleting handle.seek(pos) or
handle.truncate() from _write_buf failed no test.

The rollback's failure was covered
(test_write_buf_deletes_partial_on_non_retryable_rollback_failure). Its success — that a
failed attempt is discarded and the retry writes over clean state — was not. So the two lines
that guarantee a consistent output file after a transient write error were free to disappear
unnoticed.

The test

The first attempt puts more bytes into the stream than the retry will, then fails with a
retryable errno. That setup makes both halves of the rollback load-bearing:

• Without seek(pos), truncate() cuts at the position left after the debris, so the retried
  line lands after it.
• Without truncate(), seek(pos) rewinds but the retried line is shorter than the debris, so
  the debris tail survives past the end of the line.

In both cases the writer reports success over a malformed file — which is why the assertion is
on file contents, not on the call sequence.

It bites

Mutant	Before this PR	After
handle.truncate() → pass	survived	1 failed
handle.seek(pos) → pass	survived	1 failed
both → pass	survived	1 failed

Each fails on this test's own assertion (Rollback must discard the failed attempt's bytes),
not incidentally somewhere else — checked, because a mutant caught by an unrelated test would
not actually mean this path is covered.

Checks

Check	Result
pytest tests/test_incremental_writer.py	30 passed
pytest tests/	2243 passed, 10 skipped, 3 xfailed
ruff check / ruff format --check	clean
pyrefly baseline	unchanged at 77 — this adds none

Test-only; no production change.