client: increase default latency probe interval from 30s to 5m

[ben-dz]

Summary of Changes

• Increase the default -probe-interval for doublezerod from 30s to 300s (5 minutes), reducing steady-state ICMP control-plane load on DZDs by ~10x. With ~1700 clients each pinging ~90 devices, the old 30s interval contributed ~170+ ICMP pps per device — enough to compete with TWAMP telemetry packets under Arista COPP rate-limiting, causing spurious asymmetric link-down events.
• When the first probe finds no reachable devices (e.g. daemon starts before network is ready), retry at a fast interval (min of configured interval and 30s) until a device responds, then switch to the steady-state interval. The first probe still fires immediately on startup, and probeReady is set unconditionally after the first pass so the CLI can proceed.

Diff Breakdown

Category	Files	Lines (+/-)	Net
Core logic	2	+25 / -2	+23
Tests	1	+79 / -0	+79
Docs	1	+5 / -0	+5
Generated	1	+2 / -0	+2

Mostly test coverage for the new fast-retry behavior; core logic change is compact.

Key files (click to expand)

• client/doublezerod/internal/latency/manager_test.go — new test verifying fast-retry when first probe finds no reachable devices, and that probeReady is set regardless
• client/doublezerod/internal/latency/manager.go — add converged flag and hasReachable() check to control fast/slow probe interval; add maxInitialProbeInterval constant
• client/doublezerod/cmd/doublezerod/main.go — change -probe-interval default from 30 to 300

Testing Verification

• New unit test TestLatencyManager_FastRetryWhenUnreachable verifies: probeReady is set after first probe even when all unreachable; second probe fires at the fast interval (~30s) rather than the steady-state interval (1h in test); probe count advances as expected
• All existing latency manager tests pass (14/14)
• E2E tests unaffected — client entrypoint hardcodes -probe-interval 5
• Operators can still override with -probe-interval <seconds>