Avoid worse-case performance issues when computing test diffs
Avoids the effectively infinite diffing observed by @graven in lhcb/LHCb#252 (closed).
It's not quite equivilent as it doesn't produce the ?
lines like:
- HltANNSvc WARNING key 0 has an explicitly configured overrule -- using that...
+ HltANNSvc WARNING key 0x######## has an explicitly configured overrule -- using that...
? +++++++++
Another option would be to use difflib.unified_diff
or difflib.context_diff
but that changes the output even more.