move link_wav.py to paddlespeech/t2s/exps/gan_vocoder/ so that different vocoders can use one script.

add --dataset and --rootdir to voc3 finetune script, also remove --speaker-dict in voc5.