High-resolution timing infrastructure

[codingwithmagga]

Description

Implement a timing layer using std::chrono high-resolution clock to measure kernel execution duration accurately.

Acceptance Criteria

• Start and stop timestamps are captured with sub-microsecond resolution
• Elapsed time is exposed in multiple units (ns, µs, ms)
• Timer overhead is measurable and documented
• Timer is safe to use across multiple threads independently

Notes

TDD — tests are written before implementation code.