Fix handling of timeseries with 0 to 3 points in Vectorization

[sebhaub]

• Fix ArrayOutOfBoundsException when a timeseries has 0/1 points (happened in line 100)
• If a timeseries has only 0..3 points, previously an empty timeseries has been returned (since use_point only contains 0´s). Now in such a case the original timeseries will be returned.
• refactored Vectorization.compute() method (pre initialize use_point array, remove unneeded for-loops, reduce nesting)
• add method overload for transform with additional tolerance argument