The original length of pts made sense because when wantPos is a decimal position like 3.5, you want the window for quality 1 to contain positions 3 and 4. Having a window of 1 more doesn't make sense for that. My previous reasoning for increasing the window size was arguing from the point of wantPosition always being an integer value, which isn't the case.

Two bugs:
- SetRatio() could cause out of range panic when the ratio was sufficiently big. This happens when the newly calculated position was casted to an int. This caused the value to be rounded down. If difference times the ratio was bigger than the buffer, the calculated wantPos could be a negative index in buf1.
- When the ratio is sufficiently big, the Stream function may not only need to fill the buffers with new data once, but multiple times. The if statement has been replaced with a for loop, so the check is executed until the desired samples are reached.