Invalid GTF entries are not handled gracefully even with `ignore_bad`

[JackCurragh]

As raised in #138 there are GTFs produced by RefSeq that have ### as their last line.

Ideally pr.read_gtf(gtf, ignore_bad=True) would handle this and skip that problematic line but saving that a more informative error message could be raised directing the user to remove the offending line.

Again, very happy to submit a PR for this if you wish!

[endrebak]

Thanks for offering. I will mull over how to best do it

[endrebak]

So it seems there is no speed penalty at all for using comment="#" in pandas.read_table. So we should use it to fix this issue.

This means we can remove the custom parsing to find the comment lines in the beginning of the file also. See https://github.com/pyranges/pyranges_1.x/blob/f3d21841d8f517d3767b57876fa08acccb77490f/pyranges1/readers.py#L352

[endrebak]

Can you try again now? Also please check if there is a noticeable speedup. Just curious

[endrebak]

@marco-mariotti reading gtfs should be 50% faster now