Using a plain waveform editor would be a good idea if you were dealing with a 'pure' tone — essentially, a single sine wave, or perhaps a small number of sine waves such as used in DTMF. The sound of any human speech vowel is more complicated, and the kind of operations that are simple on its waveform tend to introduce artefacts (which you may or may not care about.) Consider that a vowel comprises of many frequencies, with slightly different wavelengths, and it's very difficult to cut out a piece of waveform that would contain a whole number of the waves you care about, and if you get a fraction of any of them, an audible crackle is likely to happen at the point the ends of the looping window are joined.
Typically, if you want to stretch a complex sound like that while reducing the editing artefacts, you'll want to start by splitting the original sound into its frequency spectrum using Fourier transformation. The selected part of this can then be mechanically stretched (or alternatively, shrunk, if you wish) by repeating some or all frames a few times, as dictated by the factor of stretch, or skipping some frames according to the factor of shrink, and the modified spectral form then re-synthesised into a new waveform. This way, the individual frequencies of different wavelengths in the loop window will end being re-generated, and it won't matter that the length of the loop window was not a whole number of some (or, well, almost all) of these waves. You can think of the process as adjusting the phases of these waves separately at each clipping window rollover point.
Some sound editing packages include tools for filtering out crackling artefacts. You can generally assume that these will be performing Fourier transformation behind the scenes, detect short high-volume events and suppress them in the spectrum, and then re-synthesise the waveform.