Removing the selectors for in situ expressions

Hi all,

During the latest teleconference we discussed in length the pros and cons of keeping/removing the selectors for in situ constructs,
and we had a consensus that, overall, it would be probably a good thing to remove them.

- Nothing would change for the dislocated rules.
- The in situ rules would be now limited to just the attributes carrying the data category information, and would never have
selectors.

The main drawback we found was that assigning ITS data categories to attributes would be more difficult as this could be done now
only through dislocated rules. This would affect only cases when you are overriding information for attributes, which, we thought,
would be rare enough to not be an issue.

Some of the advantages we found were:

- easier to validate the XPath expressions (now that they would be all dislocated)
- easier implementation
- make the precedence rules more clear
- less complicated to understand for the users
- this would resolve resolve the open issues 2620, 2621, and 2719.

We've decided to make the changes in the different test implementations we have to test this modifications (as well as the potential
its:ns element we discussed separately).

If anyone as objections to not having selectors for in situ rules, please post them as soon as possible. Otherwise, Friday Felix
will start changing the specifications to reflect the modification.

Cheers,
-yves