Add benchmarks for the geosearch

[irevoire]

closes #336

Should I merge this PR in #322 and then we merge everything in main or should we wait for #322 to be merged and then merge this one in main later?

[irevoire]

Oops I forgot to update the README and the bench.yaml. I'm putting this in draft PR for now sorry

[curquiza]

Thank you Tamo!!

[irevoire]

@Kerollmops, should I merge this in the #322, or should we first merge the geosearch and then the benchmarks on main? What do you prefer?

[Kerollmops]

I prefer that we merge #322 and this one after :)

[irevoire]

Ok so I guess we can merge now 😁

[Kerollmops]

Thank you!
Have you tested it with the CI, already?

[irevoire]

Yes!
But now I'm wondering, maybe we could add a bench on a really small dataset? ~10.000 documents

[Kerollmops]

How much time does it take to run this one? If it is much longer than the other ones, maybe we should indeed!

[irevoire]

No, currently, all the bench related to the geosearch are super fast. It takes 20 minutes for the search and an insignificant amount of time in the indexing benchmark (1h50m5s before the geosearch vs 1h50m6s with this branch).

So I think it would be good to keep this dataset of 1M documents because it's what we want to improve.
But at the same time, add a really small dataset to ensure we don't lose any performance on small datasets.

[Kerollmops]

You can open a new PR to add a small dataset if you want, I prefer to merge this one for now :)
bors merge

[bors]

Build succeeded:

• Run Rustfmt
• Tests on macos-latest with stable
• Tests on ubuntu-18.04 with stable
• Tests on windows-latest with stable