[research] Add a new problem of nbody simulation

[CharlieFRuan]

We add a new problem of nbody simulation. There are two subproblems which only differ in the number of particles, 10k and 100k.

10k (5.5x speedup gives 100):

• human solution is 2.737x
• cursor-optimized with Opus 4.5 with 5 iterations is 4.55x

100k (7.5x speedup gives 100):

• human solution is 3.935x
• cursor-optimized with Opus 4.5 with 5 iterations is 6.049x

The above numbers are based on a c7i.4xlarge

The idea of the project is modeled after https://github.com/cmu15418f24/asst3 with modifications.

Testing done with:

• Be under either random_10k/ or random_100k/ and run bash run_evaluator.sh /home/ubuntu/Frontier-CS/research/solutions/nbody_simulation/random_10k/reference_baseline.cpp