[pranavvadd]

Summary

Makes the simulation time step work end to end. RaceConfig.time_step_minutes already
existed, but weather data was always generated at 1-minute spacing, so any other step
gave a wrong time axis and wrong distance/SoC. Example: with a 5-minute step the race
ended after about 29 minutes of race time (should be about 160) and covered 142 km
instead of 196 km.

Changes

• sim/weather.py: new resample_to_time_step(), applied to both synthetic and API
  weather. Steps over 1 minute average the data (preserves solar energy), steps under
  1 minute interpolate, and a 1-minute step is unchanged.
• sim/main.py: new --time-step-minutes option (default 1.0, must be a finite number
  above 0). It replaces the three hardcoded 1.0 values and works for single runs,
  --plan and --speed-sweep.
• README.md: documents the flag.
• tests/test_time_step.py: 158 new pytest tests.

Testing

python3 -m pytest tests/test_time_step.py passes. After the fix, distance is 196.2-196.4 km
at every step from 0.5 to 10 minutes. The tests also fail (48 failures) if the old behavior is
restored. They cover resampling, synthetic and API weather, the simulator against a
1-minute baseline, all four strategies, interval-hold, and the CLI.

Related Issue

Closes #8

Notes

• Adaptive strategies (pi, stepped, interval-hold) react per step, so coarse steps
  aren't identical to the 1-minute run. The tests allow up to 10% distance difference.
• Add CLI Option To Run Simulator Without Weather API #46 (--no-api-weather) also changes sim/main.py, in different lines. Whichever of the
  two PRs merges second may need a small conflict fix.

[Kamal-uva]

yeah I merged the other one first, so small change for that and another small bug fix, the commits u make on the branch will show up here so we can marge once its handled:

Overall looks good! The test coverage is especially thorough.

One edge case I noticed in resample_to_time_step(): when the time step doesn't evenly divide the race duration (e.g., 7 minutes for a 480-minute race), the final step is treated as a full interval. Since the simulator uses the full dt_minutes for every step, the race can run past its intended end time and overestimate distance and energy consumption.

Could we handle the final partial step and add a test to verify the simulated race duration stays correct for non-divisible time steps?

Otherwise, good