refactor: change synthesize api to only specify the inputs (i.e. text or spec)
this is because the text-to-spec model synthesize command is actually also capable of synthesizing wav
this commit also clarifies what inputs are needed for copy synthesis
Fixes #182
