Add MoE route-trace replay and timing validation

[Happymic]

What This PR Does

This PR adds the first route-trace replay harness for MoE timing. It lets us feed a deterministic routing trace into the emulator, run the corresponding workload, and check the resulting timing/byte accounting.

Concrete Changes

• Adds a route-trace schema and validator.
• Adds a sample trace generator for small deterministic timing tests.
• Adds a replay runner that consumes trace data and drives emulator workloads.
• Adds export utilities for timing results.
• Adds validation gates for cycle totals, HBM bytes, overlap assumptions, and timing-stack sanity checks.
• Adds an opt-in prefetch/compute overlap timing path plus an overlap golden workload.
• Adds a portable timing smoke script that uses PLENA_VENV, PLENA_PYTHON, or the current Python instead of a hardcoded local path.

Why This Is Separate

This PR proves the measurement workflow on controlled traces before connecting it to Qwen real-routing data. It is the bridge between the timing core in #93 and the Qwen-specific replay in #95.

Validation

Ran locally before the portability cleanup:

nix develop -c bash -lc 'cd transactional_emulator && cargo build --release && cargo test --workspace --release'

export PLENA_ROOT=/path/to/plena
export PYTHONPATH="$PWD:$PWD/PLENA_Compiler:$PWD/PLENA_Tools:$PWD/transactional_emulator/testbench${PYTHONPATH:+:$PYTHONPATH}"
python3 transactional_emulator/testbench/timing_goldens/check_timing_goldens.py --plena-root "$PLENA_ROOT"
python3 transactional_emulator/testbench/timing_goldens/check_timing_goldens.py \
  --fixture transactional_emulator/testbench/timing_goldens/overlap_golden_workloads.json \
  --plena-root "$PLENA_ROOT"

After removing the hardcoded local venv path:

bash -n transactional_emulator/testbench/timing_goldens/run_timing_smoke.sh