For YT, processing separate videos greatly increases extract_info time,
but unnecessary when obtaining only track title.
For Yandex, processing is much faster, and we NEED it to get title.
Small note on how it works, copied from
raise_on_irrelevant_result() docstring:
Raises ValueError
if no words from track title are present in search result track title
and no words from artist name are present in search result artist name
I've misunderstood the FS limits, 255 is for filename, not path.
Path is limited only by libc, it's 4096.
The default outtmpl in musicdlp contains slashes, so it's a path.
But YDL considers it to be a filename, so the whole
outtmpl formatting result is trimmed to path_length.
Do we really need this? I think there are no "malicious" long-named tracks :)