Don't define typedef-names in the library wording

[frederick-vs-ja]

Since (non-template) typedef-names are declared but not defined ([basic.def]/2.9, [basic.def]/2.10).

• In [library], I think sometime it would be better to use "shown" or "specified".
• The member type value_type (or sometimes type?) is not always a typedef-name, but it is in most case. So, I'm changing to use "declared" for it.
• For member types iterator_category and iterator_concept, this PR used to use "determined" instead of "defined" when the aliased type is talked. Now the changes to "determined" are reverted.

[jensmaurer]

I'm very much on board with replacing "defined" with "declared" when we're talking about the introduction of a name, such as "X is declared if and only if..." I think that's a straightforward fix and actually improves the presentation quite a bit.

For situations where we describe what the "value" of a typedef is, I like "defined" a bit better, because it's alluding to the mathematical idea of "definition", and that's what we do in those contexts. A better way of phrasing this without using the term "defined" might exist, but "determined" leaves open a bit who determines what. Maybe "specified to be" would work; I'm not sure.

Anyway, if you want to extract the changes of the first category into a separate pull request, where we actually talk about the (presence of) a typedef declaration, then that could probably progress faster.

[frederick-vs-ja]

@jensmaurer Now the changes to "determined" are reverted.

[frederick-vs-ja]

Rebased & force-pushed. @tkoeppe @jensmaurer