Adding config files for remaining (relevant) ann-benchmarks million-scale datasets

[cjnolet]

Also exposing new dataset_memtype option for the RAFT index types that allows us to specify (in the config files) for each index whether we want the dataset to be fed in from "host", "device" or "mmap" (host memory-mapped).

[divyegala]

Approved, one comment regarding docs clarity

[divyegala]

There's a section in the documentation that describes how to form the configuration for an index. Could you add the new memory type key there?

[cjnolet]

Discussed offline- we're going to improve the docs for the benchmarks further by specifying each of the possible indexing, build, and search arguments in a table and further explaining each of them.

[cjnolet]

/merge