Let me try to answer your questions at least in part.  My apologies for references I've missed.  For an overview of the ideas without references, you might enjoy [Pavel Safranov's talks](https://www.msri.org/people/25405) at the intro conference at MSRI.

1.What *exactly* do you mean by Dijkgraaf-Witten?  If you asked *me* what Dijkgraaf-Witten was (and this is more a statement about myself than about mathematical definitions) I would have said that it was Crane-Yetter theory for a pointed braided tensor category Vec(A,q).  Without this I can't answer your questions.

2.1 This is certainly "known" but I'm having a little trouble tracking down the exact reference.  I think [Freed-Hopkins-Lurie-Teleman](https://arxiv.org/abs/0905.0731) for finite groups is doing what you want, based on earlier work of [Freed](https://arxiv.org/abs/0905.0731).

2.2 I really don't understand what you're asking here.  If Dijkgraaf-Witten is a special case of fully extended TFTs, then *by the definition of generalization* fully extended TFTs are generalized Dijkgraaf-Witten...

3.1 Crane-Yetter attached to a modular tensor category C is a fully local 4-dimensional TFT which assigns C itself to the point.  This is "known" to experts, but not fully in print anywhere.  There is an incomplete construction given in unfinished notes by [Walker](https://canyon23.net/math/tc.pdf), this uses somewhat non-standard definitions for higher categories and TFTs so even when its completed one can argue about whether it translates into other definitions.  There's also some unfinished work of Freed-Teleman in this direction.  The 012-dimensional part of Crane-Yetter is constructed and computed in [work](https://arxiv.org/abs/1501.04652) of [Ben-Zvi-Brochier-Jordan](https://arxiv.org/abs/1606.04769) building on Ayala-Francis's factorization homology.  I have work with [Brochier-Jordan](https://arxiv.org/abs/1804.07538) and also with us and [Safranov](https://arxiv.org/abs/2003.13812) showing that non-degenerate braided fusion categories are fully dualizable and actually invertible, and thus by the cobordism hypothesis give *framed* fully local TFTs which we think of as a framed version of Crane-Yetter, but we don't have serious calculations of what this theory yields in high dimensions.  Furthermore, as Arun points out in comments, in order to turn this into the usual Crane-Yetter you need to understand how the ribbon structure on a modular tensor category gives an SO(4)-fixed point structure.  You can see some results in this direction in my MSRI talk on joint work in progress with [Douglas-Schommer-Pries](https://www.msri.org/workshops/917/schedules/28167), but we're still a ways off from giving a full answer.

3.2 A good reference for how to get a state-sum out of a fully extended theory is [Orit Davidovich's PhD thesis](https://repositories.lib.utexas.edu/handle/2152/ETD-UT-2011-05-3139).  In principal there's no problem doing this, but in practice there's plenty of interesting questions along these lines to which we don't yet know the answer.  For example, you should be able to use my work with [Douglas-Schommer-Pries](https://arxiv.org/abs/1804.07538) to get a *framed* Turaev-Viro state sum model attached to a (not necessarily spherical) fusion category.  But we don't even have a guess for what exactly that should look like.  Or, after you've analyzed the SO(3)-fixed points enough, you should be able to show that the TFTs coming from spherical fusion categories via the oriented version of the cobordism hypothesis and our work agrees with Turaev-Viro thereby showing that Turaev-Viro theories are fully extended.  I expect within 5-10 years we will have all of this understood, but we don't yet, and there may be other approaches that are more direct.

Two other helpful references (with extensive bibliographies) which I didn't mention specifically are recent papers by [Schommer-Pries about invertible TFTs](https://arxiv.org/abs/1712.08029) and [Reutter's paper on semisimple theories](https://arxiv.org/abs/2001.02288).