[Bug] I-BNN tutorial benchmarks need updating

[slishak-PX]

🐛 Bug

In the documentation for BoTorch 0.11.3, I-BNN is shown to outperform standard GPs with Matern and RBF kernels:
https://botorch.org/v/0.11.3/tutorials/ibnn_bo

In the 0.12.0 documentation, the Matern kernel's performance increases substantially and dominates the others, contradicting the text:
https://botorch.org/v/0.12.0/tutorials/ibnn_bo

I initially assumed this was due to the dimension-dependent LogNormal prior, but it doesn't look like this is used in the tutorial.

[saitcakmak]

Thanks for flagging. Looks like it was switch to LogEI #2499 that helped Matern kernel. I generally dislike such comparisons in tutorials. We need tutorials to be runnable with little compute for the CI to run within a reasonable time, and comparing multiple methods on a single replication just produces unreliable results.

[Balandat]

Interesting. Impressive improvement from logEI, but I agree with @saitcakmak that these performance comparisons aren't the best fit for the tutorials. I think if we do have them then we should be setting them up in a way that can reliably produce results in a full (non- smoke test) run. Everything else should belong into some paper or more comprehensive benchmark.

cc @sdaulton, @SebastianAment

[slishak-PX]

I agree with the points about removing benchmarks from tutorials unless they're fully reproducible. However, with the diverse set of models in BoTorch that individually claim to perform well at high dimensional optimisation (e.g. SAASBO as another example), it would be great to tie them all together somewhere (whether that's in the docs or in a paper) and have it prominently displayed in the docs!