Updated the support of association filter to expect operators

[abellotti]

While this worked earlier, it was an ambiguous specification when no operators are present, i.e. filter[attr]=value case where eq is not specified.

• Updated the support of association filter so that specification of associations in filter requires the operator to be specified.
• Implied "eq" for values is no longer supported for associations, the "eq" operator should be specified.
• Updated README.md for additional examples.

[bdunne]

I think my only concern is that we're not allowing the implied eq comparator if no comparator is given and that is inconsistent between direct attributes and association attributes

[abellotti]

If we want to allow the implied eq with associations (which is what we have today), then the change instead would be to the specs in topo land for [attr][bad_op]=value would be seen as attr.bad_op = value instead of attr bad_op value, so changing the expected error message seen. (i.e. bad association attr instead of bad operator bad_op). Let me know @bdunne.

[abellotti]

@lindgrenj6 @syncrou any preference here ?

[syncrou]

@abellotti so are you saying the implied eq for associations is only a spec change in topo?

[abellotti]

correct, if we want implied eq, it's already here in common, so I'd close this PR and update specs in topo instead.

[abellotti]

any update on this ? should we stick with the optional eq for associations ?

[lindgrenj6]

@abellotti I'm of the opinion maybe we should still allow implied eq, but that's just me. Thoughts @bdunne @syncrou?

[abellotti]

Closing this PR in preference for supporting the implied eq on association filters as Sources is already GA'd with this capability.