change how textgrid2rttm works to interpret interval contents

[alecristia]

assume the content of the interval is "what the speaker said", except for CHI, where we first check for the VCM categories, and otherwise assume it's content