Doc2doc search to improve index performance on hard queries.

[khaiwang]

• add compile command in CMAKELSITS
• data.ipynb is a demo to show the recall distribution of the query set
• buildtext2image10m.ini is the build config for text2image ood dataset, see https://github.com/harsha-simhadri/big-ann-benchmarks neurips23 ood track for dataset information
• Set following args to use doc2doc in a build config file (see buildtext2image10m as an example):
  • spreadSearch is the option for doc2doc, spreadSearch = true will enable doc2doc
  • doc2docRounds can decide the max doc2doc search round
  • doc2docResults is the max number of doc used for doc2doc search in each round