WL#2565: support relative column weights in the index

one should be able to specify relative importance of fields when creaking an index.

e.g. one could create a combined index (title, keywords, body),
specifying that a matck in the title is 5 time better (makes a document 5 time
more relevant) than a similar match in the body, and a match in keywords is 50
times better (as keywords are usually hand-picked to represent document's content)

the syntax could be, for example

  FULLTEXT(title*5, keywords*50, body)