You can't do that with AVAssetExportSession because the presets are quite fixed in their usage. A preset value of AVAssetExportPresetPassthrough will keep your input formats on output.
As your task will be manipulating the audio sample buffers directly you should use the second variant that AVFoundation will give you: paired AVAssetReader and AVAssetWriter setup. You'll find proper sample code as in AVReaderWriterOSX from Apple developer source. This should also work with iOS besides you have different I/O format settings available. The availability to decompress audio as PCM and write back to uncompressed .wav file should be given.