test(client/doublezero): return different min, max, avg from make_latency

[axaysagathiya]

Summary of Changes

Updated make_latency function in tests to accept and return distinct latency values for minimum, average, and maximum latency fields.

Closes #3495

Testing Verification

cd client/doublezero && cargo test dzd_latency

[axaysagathiya]

Is the current approach acceptable, or do we want to modify it?

We can accept the min_latency_ns as a function argument.
Then, we can calculate
the avg_latency_ns as min_latency_ns plus 1 millisecond, and
the max_latency_ns as min_latency_ns plus 2 milliseconds.

[martinsander00]

Thanks for picking this up! Two suggestions to make the change worth it:

1. Convert the two existing "min vs avg" tests to use make_latency, they're natural callers for the new signature:

• test_retrieve_latencies_sorts_by_min_not_avg (lines 1013–1032)
• test_best_latency_current_device_seeded_by_min_not_avg (lines 1066–1085)

2. Make one existing test adversarial. Right now every call site passes monotonic triples (10M, 11M, 12M), so swapping min_latency_ns → avg_latency_ns in the ranking code wouldn't fail anything. In test_best_latency_selects_lowest, change the values so min and avg disagree on the winner (e.g., pk1 min=12M/avg=5M, pk2 min=9M/avg=20M, pk3 min=15M/avg=15M), same assertion, but now it actually catches a wrong-field bug.

[martinsander00]

two things:

1. can you use the same number notation we use in the rest of the codebase? Rn looks like you are using two different notations (1_000_000 vs 1000000)
2. seems like due to some cross repo access policies our workflows don't run on forks. I'll chat with the team to see what we can do about it