Use Eval instead of Trampoline for State

[ceedubs]

This may provide the following benefits:

1. It should be a bit more efficient.
2. It removes the need to import cats.std.function._ for some
   operations, which may make it a bit more newcomer friendly.
3. We are already using Eval for things like foldRight, so it seems more
   consistent to use it for State.
4. I think it's a bit easier to explain Eval than Trampoline, since
   it's hard to explain the latter without first explaining Free. So
   again, this may be a bit more newcomer friendly.

However, there may be downsides I haven't thought of. I thought I would throw this PR together for discussion. Thoughts?

[ceedubs]

Paging @refried @non @jdegoes @adelbertc @stew @mpilquist as people who have shown interest in State and/or Eval previously.

[codecov-io]

Current coverage is 88.30%

Merging #782 into master will decrease coverage by -0.18% as of 8d35ddb

@@            master   #782   diff @@
=====================================
  Files          166    166       
  Stmts         2292   2291     -1
  Branches        75     75       
  Methods          0      0       
=====================================
- Hit           2028   2023     -5
  Partial          0      0       
- Missed         264    268     +4

Review entire Coverage Diff as of 8d35ddb

Powered by Codecov. Updated on successful CI builds.

[milessabin]

👍

I found working with Eval in Kittens a lot more straightforward than using Trampoline.

[ceedubs]

I noticed there was an unused instance specialized to State. It may have been necessary before this change - I'm not sure - but scalac happily picks up the StateT version, so I've removed it which should fix the reported decrease in code coverage from this PR.

[adelbertc]

👍