Revise handling for make check with multiple log files

[jimklimov]

Recently support was added to stash all test programs' logs and traces after a make check, however if any test fails, the best we currently can do is mark them all failed (which is a bit irrelevant when you have 10 "failed" links and some one is propagated to github, but only one or two really contain an error trace).

This issue is to suggest some way out, e.g.:

• analyze TAP logs to see who failed and report better targeted info, or
• make a tarball of all those logs we report now and still have many irrelevant logs delivered, but all under one URL in reports so at least searching for the haystack with the needle is easier.