[Feature]: Add Time Consumption Evaluation for Key Operations

[magicbrighter]

Description:

To better understand and optimize performance, we should add explicit time consumption (latency) measurements for critical operations in the codebase. This will help identify bottlenecks and track performance regressions over time.

Suggested Implementation:

• Instrument key functions or workflows with high-resolution timers.

Example Use Case:
When processing a batch of data, it would be valuable to know how long each stage (loading, transformation, saving) takes.

Acceptance Criteria:

• Time measurement added to at least one core operation.
• Results are clearly logged or accessible.
• Minimal overhead introduced by the timing instrumentation.

Let me know if you'd like this scoped to a specific module or feature!

[helloml0326]

Hi there, thanks for the proposal! Adding explicit latency measurements is definitely valuable for our performance tracking and regression testing.

Regarding the implementation, to keep the codebase clean and minimize overhead, I suggest introducing a lightweight decorator (or context manager) utility (e.g. inside utils/timer.py).

For example:

It would be great if this utility could integrate seamlessly with our existing logging system (e.g., logging at DEBUG level or aggregated in a summary) so it doesn't clutter the standard output.

Scope: Feel free to start with the core evaluation workflow or model inference methods as a pilot. We'd love to see a PR from you to implement this!

[Alexxigang]

I traced the core evaluation path locally and I'm working on a pilot implementation for this in the grading workflow. The plan is to add a lightweight reusable timing utility under openjudge/utils/timer.py, instrument GradingRunner with optional latency collection/logging, and add regression tests for both the timer utility and timing summaries exposed by the runner.

[Alexxigang]

I opened a PR for this here: #160.