Update links to point to the Internet Archive for issue2799 and issue5726

[timvandermeij]

We remove yo01. The file is not available, it was introduced in 3529658 along with another test file (so we should not need it) and it was replaced in 6b2c6fc, so it was only in the repository for a very short time.

Fixes #6854 and #5091. All test cases are now either reduced or link to trusted sources like Bugzilla and the Internet Archive, making sure that they keep on existing and do not change. MD5 mismatches have also been resolved.

[Snuffleupagus]

We remove yo01. The file is not available, it was introduced in 3529658 along with another test file (so we should not need it) and it was replaced in 6b2c6fc, so it was only in the repository for a very short time.

I'm not sure if I agree with this conclusion, since I see no indication that usmanm-bad-auto-fetch was intended to replace yo01. It rather seems to me that it's more likely that, given that it was already unavailable at the time of that patch, it might have been removed a little to hastily.

I'm assuming that the two files added in 3529658 doesn't demonstrate entirely identical issues, since otherwise I cannot image why both were added. @timvandermeij Did you verify that both those files failed in the same way previously?
(I've been meaning to replace yo01 with a reduced test-case, but I've not yet had the time).

[timvandermeij]

I should have been a bit more clear indeed. I tested both files from 3529658 (the unavailable file substituted by https://web.archive.org/web/20120622065546/http://www.ugokotsu.co.jp/ug/rosen/jikoku/yo01.pdf) with the previews from #2839 (just before the fix, both broken because characters are shifted to the right) and #2848 (just after the fix, both fixed because the characters are in the right position). That is how I came to my conclusion, but it could be that I have missed something that you do notice! In that case, please let me know. We can then do two things: either remove the file, or I'll leave yo01 out of this commit so you can replace it with a reduced file. I'm fine with either way!

[Snuffleupagus]

OK, I've just tested this using the "right" version of yo01.pdf (I still have a copy locally), and it appears that the failures are the same as with 20130226130259.pdf.
Looking at issue #2842, it specifically mentions Windows XP (which I'm not using), but it does seem highly likely that those two files demonstrate the same issue. And I suppose you do have a good point about that test being very short-lived!

So in summary, I'm fine with just removing the link (mean less work ;-), so I'll test/verify this locally later tonight and then merge it. Thanks for providing additional context here!

[Snuffleupagus]

Looks good, thank you!