Using symbols in an =in=> value causes test to incorrectly succeed

```
(deftest reproducer
(facts "This correctly fails"
{:foo [1 2]} =in=> {:foo [1 3]})
(let [sym :foo]
(facts "This should fail, but succeeds"
{sym [1 2]} =in=> {sym [1 3]})))
```

Start by running the supplied `deftest reproducer`, comparing the literal `:foo` case with the `sym` binding in the `facts` assertions. Trace how `=in=>` handles map keys and add a regression test so the symbol-key case fails correctly rather than succeeding.

Written by the indexing model from the issue text.