TextSplitter creates incorrect bound checks.

The TextSplitter implementation does not advance the start point for calculating the subsequent split. For example to have three splits between a and d, the splitter should produce splits like ( col >= 'a', col < 'b'), (col >= 'b', col < 'c') and (col >= 'c', col <= 'd'). However, what it produces instead is ( col >= 'a', col < 'b'), (col >= 'a', col < 'c') and (col >= 'a', col <= 'd'). This results in intersecting splits which causes duplicate records.