test(node-js): add unit tests for controller command logic; fix DriveValue.read rounding

[Arvin7liu]

Summary

The node-js web controller has no automated tests for its server-side command logic, even though that logic computes the exact left/right motor commands sent to the robot. This PR adds dependency-free unit tests and fixes a latent rounding bug found while writing them.

• New tests (controller/node-js/test/commands.test.js, run via Node's built-in node:test):
  • CommandHandler drive mapping for goForward / goBackward / forwardLeft / forwardRight / rotateLeft / rotateRight / reset.
  • DriveCommandReducer de-duplication (identical consecutive drive commands are not re-sent).
  • CommandHandler.sendCommand raw-command forwarding.
  • DriveValue clamping to [-1, 1] and read() precision.
• Bug fix: DriveValue.read() used Math.round(value, 3). Math.round only accepts one argument, so the 3 was ignored and the method returned an integer instead of a 3-decimal value. Changed to Math.round(value * 1000) / 1000. DriveValue is now exported so the tests can reach it. Note: read() isn't currently called anywhere in server/ or client/ (the drive-command path uses the raw .min()/.max()/.write() values directly), so this fix doesn't change any live behavior today, it's a correctness fix to a previously-unused method, now exercised by this PR's own tests.
• Added a "test": "node --test" script to controller/node-js/package.json (zero new dependencies; runs on Node 18+).

Why this matters

The CommandHandler drive mapping tested here is the code that computes the actual left/right motor commands sent to the robot, so a regression there would silently send wrong motor commands to a physical robot. The new tests lock in that expected behavior and run in CI-friendly, dependency-free fashion.

Test plan

• cd controller/node-js && npm test → 10/10 passing on Node 22 (also verified with node --check on all server/*.js and client/*.js).
• Regression check: reverting the read() fix makes the DriveValue test fail (proving the test actually guards the fix).

Notes

This is unrelated to #447 / PR #514 (the TF/Keras policy-training issue), which is handled separately.

🤖 Generated with WorkBuddy

[thias15]

Nice, dependency-free test coverage; traced the assertions by hand against CommandHandler/DriveValue and they're all correct, including the negative-rounding case.

One correction: DriveValue.read() isn't actually called anywhere in server/ or client/ today. sendDriveCommand uses the raw .min()/.max()/.write() values directly, not .read(). So the rounding fix wasn't live-affecting motor commands, it's a fix to a previously-unused method now exercised by this PR's own tests. I'll update the description to reflect that.

[thias15]

Verified all test assertions by hand against the actual CommandHandler/DriveValue logic; they're correct. CI is green. Approving.