tree-conflicts: please note r34454 and r34455

Hi tree-conflicts folks,

I changed the update_ switch_ and merge_tests.py DeepTreeTestcases so that
they expect the actual desired behaviour, not our development state in r34454.

I found and fixed a bug in update that didn't see node addition/deletion as
a local mod in r34455. A quick review would be nice.
Do we also need to check for entry->deleted?
Do we need to check "hidden" nodes?
Do we need to use svn_wc__walk_entries_and_tc()?
And is it good to include that schedule check in entry_has_local_mods(),
even though it is both exclusive and blindingly simple?

Well, in result, all of the update_ switch_ and merge_tests.py
DeepTreeTestcases justly pass except use case 5 (merge with incoming tree del).

~Neels