Clarify regression

From https://gitlab.slack.com/archives/C0NFPSFA8/p1507131184000324:

Hmm… the definition in Process.md is kind of vague. Doesn’t it have to not only be a bug in a release, but an actual loss or degradation in functionality from a previous release? Otherwise it’s just a “new bug”. (edited)
If I ship a new feature with a typo, that only shows when using that new feature, then it’s not a regression. It’s a bug. Now, it could still be a bug worth of fixing in Next Patch Release, but it’s not a regression.
...
Shipping a buggy feature is different than breaking a previously working feature.
When finding a bug in a new feature, you have to look at the severity of the bug before determining if it should be in a patch or just fixed in the next release. But when finding an actual regression, the implied severity is much higher because you’re breaking someone’s trust (and likely affecting their productivity because they started depending on the existing functionality).