* Support parsing srt without sequence numbers.
There are many srt in the wild that are missing sequence numbers. It's
easy to detect by looking for an empty line vs index/sequence in the
line before the time. In my analysis of many srt I have not seen ones
missing the empty line, but it's feasible that these also exist. This
shouldn't affect parsing otherwise, however, and in my experience the
missing sequence number is common enough that it's useful to handle.
* Remove duplicate testdata file and use existing.
