argosfs: use bounded rootfs group commits

[fwerkor]

Summary

• pin ArgosFS to b8f84e66932d805612509fc4c4ff77b5b02913ec from fs: bound deferred rootfs group commits ArgosFS#109
• enable explicit 5-second / 128-transaction bounded group commits for the root filesystem
• reduce the single-device rootfs chunk size from 256 KiB to 64 KiB
• bump the ArgosFS package release

Why

CapOS enables deferred data, metadata, and journal flushes to avoid excessive per-write barriers. The previous ArgosFS implementation left those batches unbounded. The pinned revision adds periodic and transaction-count durability boundaries, synchronous fsync/O_SYNC behavior, data-before-journal ordering, and crash-safe deferred extent reclamation.

The 64 KiB chunk also reduces the worst-case data rewritten by a 4 KiB random overwrite from 256 KiB to 64 KiB for the current k=1,m=0 root layout.

Validation

• exact ArgosFS revision is reachable from the upstream repository
• image recipe contains the expected chunk size and durability bounds
• git diff --check
• relevant CI shell syntax check
• ArgosFS revision passed 23 unit, 136 integration, and 5 raw readonly regression tests plus Clippy

Depends on fwerkor/ArgosFS#109.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.