Revisit decision to make <> optional in certain cases

[msporny]

Making <> optional in certain cases makes it difficult for people to know what to expect when they receive objects. We may benefit from stricter, more simple rules, such as:

If the object's value is an IRI, you MUST surround the value in "<>" characters.

instead of what we have in the spec right now:

http://json-ld.org/spec/ED/20101128/#iris

[msporny]

Another alternative that was proposed was:

For predicates, @ and a - if you are normalized you MUST use <>, if you are not normalized you MUST NOT use <>.