fix(quake): skip latency emulation on WSL2

[huklaa]

Fixes #31
Quake-generated latency setup scripts currently fail on WSL2 because the required tc qdisc kernel modules are unavailable. Detect WSL2 via /proc/sys/kernel/osrelease and skip latency emulation there so localdev can continue starting without manual script edits.
Normal Linux environments keep the existing tc/netem behavior unchanged.

[osr21]

The approach here is sound, and one detail worth stating explicitly for reviewers: the generated script runs inside the containers, and on Docker Desktop / WSL2 backends the container shares the WSL2 utility VM's kernel — so /proc/sys/kernel/osrelease inside the container also reports microsoft-standard-WSL2. In-container detection is therefore reliable, which isn't obvious at first glance.

Three observations from reviewing the diff (including the latest commit bfc0bfa):

1. Trailing whitespace introduced in bfc0bfa — the "add trailing newline" commit fixes the missing EOF newline in crates/quake/src/latency.rs, but the added line before Ok(node_regions) contains trailing spaces (and doubles the blank line). cargo fmt strips both, so this will trip a cargo fmt --check gate. A quick cargo fmt pass on the branch would clean it up.

2. The pattern also matches WSL1 — grep -qi microsoft matches WSL1 (4.4.0-…-Microsoft) as well as WSL2 (…-microsoft-standard-WSL2). That's arguably desirable since WSL1 lacks netem too, but the echoed message says "WSL2 detected". Either generalize the message to "WSL detected" or pin the pattern to microsoft-standard if the intent is WSL2-only.

3. Consider a capability probe (or an override) instead of platform detection — Microsoft's stock WSL2 kernel omits CONFIG_NET_SCH_NETEM, but a common documented workaround is running a custom-built kernel with netem enabled; those users would now get latency emulation silently disabled even though their kernel supports it. A probe skips only when netem is actually unavailable:

   if ! tc qdisc add dev lo root netem delay 1ms 2>/dev/null; then
     echo "netem unavailable on this kernel; skipping latency emulation."
     exit 0
   fi
   tc qdisc del dev lo root 2>/dev/null || true

   This also future-proofs against Microsoft adding netem to the stock kernel. Alternatively, keep the osrelease check but honor an escape hatch (e.g. QUAKE_FORCE_LATENCY=1). Worth considering because the skip silently changes simulation semantics — inter-region latencies collapse to ~0, so consensus timing measured on WSL2 won't be comparable to Linux runs; the probe at least keeps emulation on for kernels that can do it.

None of these block the fix — it unblocks localdev on WSL2 exactly as #31 asks.

[osr21]

Re-checked the branch at head bcc00b9 — both flagged items are resolved:

• The trailing-whitespace line before Ok(node_regions) is gone (245cbaf), and the two unrelated edits that commit briefly introduced (the /// Sydney, Australia doc comment and the /*us-w-2 */ matrix column alignment) were restored in bcc00b9. The file now diffs against main with only the intended guard block plus the EOF newline.

• Pinning the pattern to microsoft-standard is the right call, and it doesn't regress WSL1 users either: WSL1 has no Linux kernel of its own, so Docker Desktop on a WSL1-only setup falls back to the Hyper-V/linuxkit VM — whose kernel ships netem — meaning containers never actually execute under a WSL1 osrelease string. The WSL2-only match is exactly the set of environments where the qdisc modules are missing.

The capability-probe/override idea from my earlier comment remains optional future-proofing (custom WSL2 kernels with CONFIG_NET_SCH_NETEM); as it stands the PR does what #31 needs. LGTM.