Dustin Deokjae Lee

To directly show issue in the original implementation, I printed three values - `acq_fn(best_seqs[None,:])`, - `acq_fn(best_seqs[..., None]).mean()`, - `acq_fn(best_seqs[..., None][-1]` at each iteration in the original implementation [optimizers/lambo.py#L334-336](https://github.com/samuelstanton/lambo/blob/main/lambo/optimizers/lambo.py#L334-L336) as follows:...

I re-read the original code based on the case of using EHVI instead of NEHVI. According to the implementation of EHVI, I found out that - `acq_fn(seqs[None,:])` computes q-EHVI, -...

Here, I summarize three issues which i fixed in this PR. Please refer to the changes in https://github.com/samuelstanton/lambo/compare/main...badeok0716:lambo-fix:main. 1. Error in q-NEHVI calculation `q=X.shape[-2]` in [lambo/acquisitions/monte_carlo.py#L69-L89](https://github.com/samuelstanton/lambo/blob/main/lambo/acquisitions/monte_carlo.py#L69-L89) leads to wrong calculation...

I summarized three issues (the issue about q-NEHVI calculation and two additional issues) in this repo and illustrated the performance after the correction in [here](https://github.com/samuelstanton/lambo/pull/13).