dtest: add optional --max/--min per-hand timing summary

[tameware]

Report min/max ms-per-hand across batches alongside the existing average, so extremes are in the same units as Avg user/sys time.

[Copilot]

Pull request overview

Adds optional reporting of per-hand timing extremes (min/max ms-per-hand) across batches in the dtest test driver, so variability is visible alongside the existing average user/sys timing summary.

Changes:

• Extend TestTimer to track per-batch per-hand min/max for both user and system timings, and allow print_hands() to write to a caller-provided stream with optional min/max output.
• Add --min / --max CLI flags and plumb them through OptionsType into the final timing summary output.
• Add GoogleTest coverage for TestTimer batch tracking and for --min / --max argument parsing.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
library/tests/TestTimer.hpp	Adds batch min/max state + new APIs and output options for per-hand reporting.
library/tests/TestTimer.cpp	Implements batch recording/min-max tracking and optional min/max printing to a provided stream.
library/tests/testcommon.cpp	Passes new --min/--max options into timer.print_hands().
library/tests/test_timer_test.cpp	New unit tests for TestTimer::record() min/max behavior and optional reporting.
library/tests/cst.hpp	Extends OptionsType with show_min_ / show_max_ flags.
library/tests/BUILD.bazel	Adds test_timer_test and args_test Bazel test targets and excludes them from the non-gtest filegroup.
library/tests/args.hpp	Updates args module documentation for the new min/max timing options.
library/tests/args.cpp	Adds --min / --max parsing and updates token handling to avoid collisions with existing options.
library/tests/args_test.cpp	New unit tests verifying --min / --max parsing behavior.

[Copilot]

Pull request overview

Copilot reviewed 9 out of 9 changed files in this pull request and generated 3 comments.

[zzcgumn]

This makes sense. We want to avoid optimisations that look good on average but only help for the most difficult board and increase the time for most boards.