core/storage: Only mark WAL initialized after a successful header sync

[penberg]

prepare_wal_finish's sync completion callback ignored its result and
called mark_initialized() unconditionally. Because the completion
machinery runs the sync callback on failure as well as success, a failed
header fsync still flipped the in-memory initialized flag to true. After
that, prepare_wal_start short-circuits and prepare_wal_header returns
None, so the header is never re-issued before the next append, leaving
the in-memory state claiming durability the on-disk header never reached.

Guard the mark_initialized() call on res.is_ok() so a failed sync leaves
the WAL uninitialized and the header is rewritten on the next append.

Found by Aristo.

[codspeed-hq]

Merging this PR will improve performance by 19.14%

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 3 improved benchmarks
✅ 632 untouched benchmarks
⏩ 105 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	format_float_neg_infinity	1.2 µs	1 µs	+19.9%
⚡	Simulation	format_float_infinity	1.3 µs	1.1 µs	+19.33%
⚡	Simulation	format_float_zero	1,136.4 ns	961.4 ns	+18.2%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing workspace-6-2 (6fd58f4) with main (d001d54)}

Footnotes

1. 105 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩