Add a flag to create index for SCAN and CONTAINSSTRING automatically when the dataset is loaded in memory.

The current optimization for SCAN and CONTAINSSTRING generates an index for ASCII-only text columns at the first request including either SCAN or CONTAINSSTRING. The index build operation could be up to 25 seconds long and goes in timeout if it takes more time. It would be useful to be able to set a different timeout and to force the creation of the index when the dataset is loaded in memory instead of waiting for the first query. When the column is particularly critical for the reports, the first user in the morning is the only one paying the price for the index build.

Another good improvement would be the option of persisting the index as part of the build process - but it's possible that this is too expensive if the internal structure of the index does not allow to be serialized in an effective way.