cli: UART test/diagnostic commands report failure in output but exit 0 — decide pass/fail policy

[gHashTag]

Surfaced during a silent-success audit of src/cli/*.zig (the audit that produced #427, where fpga-flash flash and verify-pid exited 0 on failure). Those two were unambiguous and are fixed. Three more are design-ambiguous — they belong to test/diagnostic commands, so whether a detected failure should be a non-zero exit is a policy call, not a clear bug. Flagging for a decision rather than changing behavior unilaterally.

Instances (each reports failure but returns/exits 0)

File	Function	Failure it prints without exiting non-zero
src/cli/fpga_flash.zig	runUartTest()	⚠️ RX empty - no echo from FPGA
src/cli/uart_test.zig	runPingTest()	Success rate: 0/N (all ❌ Timeout)
src/cli/uart_test.zig	runLoopbackTest()	Result: 0/12 bytes matched (❌)

The decision

• If these are pass/fail tests (CI/scripts gate on them): a total/partial failure should exit non-zero (e.g. error.UartTestFailed when success < count, or a threshold). This matches the now-consistent fpga-flash flash/verify-pid behavior.
• If they are interactive diagnostics (human reads the μs stats and success rate): exit 0 is fine and the current output is the deliverable.

Clean rungs of the audit — fpga_dslogic.zig (❌ lines are troubleshooting hints; real errors exit(1)) and sacred_synth.zig (return error.SynthesisFailed) — need no change.

Once you pick a policy I can apply it consistently across the three.