Expand fn literals at threading macroexpand time

[tomdl89]

• I have read the Clojure etiquette and will respect it when communicating on this platform.

• I have read the developer documentation.

• This PR corresponds to an issue with a clear problem statement. (1923)

• This PR contains a test to prevent against future regressions

• I have updated the CHANGELOG.md file with a description of the addressed issue.

---

This PR takes the functionality from #2211 which wasn't implemented in #2243.
Namely the work to thread into fn literals (#() forms) so invalid names and unused values can be detected.

Before this PR, the following code would not raise a linting error:

(-> :foo #(name %))

But after it, it raises a "Function name must be simple symbol but got: :foo" error.
This makes sense because the macroexpansion of the above form looks like so:

(macroexpand '(-> :foo #(name %)))
;; => (fn* :foo [p1__23143#] (name p1__23143#))

Also, before this PR, the following code would not raise a linting error:

(->> :foo #(name %))

But after it, it raises an "Unused value" error.
This makes sense because the macroexpansion of the above form looks like so:

(macroexpand '(->> :foo #(name %)))
;; => (fn* [p1__23147#] (name p1__23147#) :foo)

The PR is split into two commits, which should be squashed:

1. Expand fn literals at threading macroexpand time
   This is the functional changes:
   a. Add separate treatment for :fn forms vs other non-lists in expand-> and expand->>
   b. Move % anaphoric argument handling earlier in analyze-fn
   c. Add fn* to the list of core-syms checked by analyze-usages2 and lint-var-usage (as this is what fn literals expand to)
2. Move expand-fn code & add tests
   I think it's clearer to move the expand-fn code and its dependencies as a separate step, to show that nothing in it is changed. Added tests & updated the changelog.

[borkdude]

It seems kind of arbitrary to expand a fn literal specifically in -> or ->> (only such that it gives a better message in some cases). I'll have to think a bit more about this.

[tomdl89]

Thanks for taking a look @borkdude. Fair point that it's arbitrary to expand fn literals only in -> and ->>. I've worked on a new approach which expands the literals earlier in the call stack: before the enclosing list is processed. This means values can be added to the fn* body by any macro which takes the fn literal as a direct child, which covers both these threading macros, and any others like cond->. Let me know if that's what you were thinking of: master...tomdl89:clj-kondo:fn-literal-earlier-expansion