When "failed CI" isn't a blocker

[Mark J Cox]

So when we have a PR in a state where any CI is marked as failing I
lump them into a single "failed CI" state.  The stale (over 30 day old
untouched) ones are shown below.  But most of them (if not all) are
not actually CI failures.  Aside from fixing the CI, one solution
would be a label "Ignore CI failure" which can be manually set by a
committer who reviews the fail and it's harmless, then I could have
the bot ignore failed CI when that label is present.  (And perhaps the
bot also resets the label if the CI gets restarted due to changes).
That would help have things in the right state bucket and can be acted
on/reminded properly.

Mark

at Sep7: failed CI  ( 19 issues, median  276  days)
    11464  days:37 [* mjc: travis timeout on one combination, could ignore]
    11327 reviewed:commented days:177 [* mjc: same]
    11288 branch: master, reviewed:commented days:180 [* mjc: same]
    11257 branch: 1.1.1, branch: master, reviewed:approved days:178 [*
mjc: same, ......]
    11151  days:189
    10797 branch: master, reviewed:commented days:150
    10556  days:276
    10465  days:293
    9926  days:355
    9603  days:355
    9155 reviewed:commented days:229
    8955 branch: 1.1.1, branch: master, reviewed:dismissed days:304
    8115 reviewed:commented days:177
    7921 reviewed:commented days:548
    7914 reviewed:approved days:607
    7380 reviewed:commented days:695
    7051 milestone:Assessed, reviewed:commented days:744
    4992 milestone:Assessed, reviewed:commented days:317
    4338 milestone:Assessed,  days:153