fix(linux): repair the Linux build (#50)

[oreofeolurin]

Fixes #50. Found by the CI in #49 on its first run.

flo does not compile on Linux, and the next release tag would have failed

release.yml runs zig build --release=safe -Dtarget=<linux>. On dev today:

$ zig build --release=safe -Dtarget=x86_64-linux
src/node/dashboard/host_stats.zig:180:24: error: root source file struct 'fs' has no member named 'openFileAbsolute'
error: 1 compilation errors

Tagging v0.1.0-dev.10 would have failed both Linux build jobs and both Docker publishes. The last green release was v0.1.0-dev.9 on 2026-05-28; the break landed a week later in 4dd595a. No CI on PRs and no release in three months, so nothing caught it.

Three Zig 0.16 removals, all Linux-only

File	Symbol	Fix
host_stats.zig:180	std.fs.openFileAbsolute	stdx.fs.readFile
host_stats.zig:186	File.readAll	(same call — open + readAll + close collapse into one)
reactor.zig:277,287	std.posix.timerfd_create / timerfd_settime	raw std.os.linux syscalls, errno mapped the way stdx.net already does

Why every local build stayed green

Zig analyses lazily. host_stats.zig reads /proc and the timerfd calls sit behind else if (comptime is_linux) — on a macOS host neither branch is ever analysed, so they can reference symbols that don't exist and nothing complains.

This is the third time this trap has bitten in this codebase, after the metrics exporter and util/http/response.zig both turning out to be dead code that had never compiled. On this project, "it builds locally" says nothing about code behind a platform or feature gate — which is the strongest argument for the CI in #49.

Worth noting stdx.fs.openFileAbsolute already existed. The stdx wrappers are there precisely to track the std API this codebase targets; host_stats.zig called std.fs directly and bypassed the convention.

Verification

zig build --release=safe -Dtarget=x86_64-linux     # fails on dev, passes here
zig build --release=safe -Dtarget=aarch64-linux    # green

I confirmed the failure against a stashed baseline rather than assuming — that's the check that establishes releases were genuinely broken, not just cross-compilation.

macOS zig build, test-unit and test-integration all still pass. Native Linux Debug is covered by #49's CI.

Note: cross-compiled Debug Linux builds still fail in checksum_hw.zig (an x86 self-hosted-backend encoder bug, and an aarch64 crc feature gate). Neither affects native Debug or release builds — CI's native x86 job compiles it fine — so it's out of scope here.

Suggested merge order

1. This PR — unbreaks Linux
2. ci: run build and the three test suites on every PR #49 (CI) — rebase, should then go green
3. Rebase fix(stream): rebuild name registry and record counts on recovery (#42 items 2, 3) #45, fix(cluster): honor [cluster] enabled, implement cluster status, explain io_uring (#42 items 5, 6, 7) #47, fix(stream): key stream metadata by namespace-qualified name (#46) #48 so CI actually checks them, then merge each