Implement nth-of-type when we are including the tagName

[eoghanmurray]

An implementation of nth-of-type #3

The idea here is to never allow the combination of a tagName + :nth-child — from the point of view that it's

• much more intuitive for humans reading it
• might be more robust
  (reasons given by @kfrederix)

[eoghanmurray]

Highlighting one unexpected change to the test results ... the following is preferred as the algo immediately cuts off if a unique match is found at a given level:

        ++.-of2·p:nth-child(2)·>·a                   (Expected)
        --.cell:nth-child(1)·>·p:nth-of-type(1)·>·a    (Actual)

Before, .cell:nth-child(1)·>·p:nth-child(2)·>·a is checked, but found to be non-unique, so algo continues exploring higher levels.

I don't know if it's desirable to fix this or not? And don't know how best to fix.

Note: up for debate, but the 'desired' selector would be .-of2 p:nth-of-type(1) > a which is unique