URL targeting on seven conditions

URL targeting decides which pages an experiment is eligible on. Seven conditions cover the shapes real sites need: equals for a single page, starts with for a section, contains for a path fragment, ends with for a suffix, and matches regex for everything a plain string cannot express.

Two of the seven are negative. Not equals and does not contain exist because "everywhere except" is a common and awkward rule to write positively, and rebuilding it out of includes is where targeting mistakes come from.

An experiment can carry more than one rule, so a test can cover several sections without needing a regular expression to hold them all. Rules are stored against the experiment and are copied when you duplicate it.

Targeting is also what the collision model reads. Two experiments whose URL rules provably cannot match the same page are disjoint, and nothing else about them needs checking.