Fix use of _items as a normal trait in PreferencesHelper and PreferencesPage

[kitchoi]

Closes #225

This PR changes the listener for synchronizing the traits on the PreferencesHelper and PreferencesPage with the preferences, such that if the helper / page has a trait with a name suffix _items defined, it would not be mistaken as the event trait for list/set/dict mutation. Otherwise it would assume it is the event trait for list/set/dict mutation, and handle it specifically. If the assumption is wrong, the change handler will raise an error, but the error will be captured in a normal run condition and the default log message will appear.

This fix is possible because one cannot have a trait called name and a trait called name_items (not an event) and expect both of them to operate normally. Consider the following traits-only tests:

import unittest
from traits.api import *

class ItemsFirst(HasTraits):
    names_items = Str()
    names = List(Str)

class ItemsSecond(HasTraits):
    names = List(Str)
    names_items = Str()

class TestItems(unittest.TestCase):
    def test_mutate_fails_items_second(self):
        obj = ItemsSecond()
        obj.names_items = "Hello"
        obj.names = []
        with self.assertRaises(TraitError):
            obj.names.append("1")

    def test_assignment_fails(self):
        obj = ItemsFirst()
        obj.names = []
        obj.names.append("1")
        with self.assertRaises(TraitError):
            obj.names_items = "Hello"

The above tests pass. If there is a list called "names" and there is a trait called "names_items", the existence of the second trait will cause a lot of unexpected behaviours on its own already.

Checklist

• Add a news fragment if this PR is news-worthy for end users. (see docs/releases/README.rst)

[aaronayres35]

Changes LGTM.
The test that sparked my concern here: https://github.com/enthought/apptools/pull/196/files#r526423040 now passes as well. (This scenario covered by the tests added here).
Also this PR description is very clear as to why this fix is possible

[aaronayres35]

I do find this comment a little confusing... We are not conditionally doing nothing as it suggests

[kitchoi]

Good point. I think I know what this means, but I agree this is not obvious at all, I will try to fix this comment.

[aaronayres35]

This is much clearer, thank you!