Minimize use of `eval_with_capture`

[brodieG]

Right now eval_with_capture is used to evaluate test expressions, but also to compare reference to test items. The latter one actually ends up being a substantial portion of the unitizer execution time, but it seems that the actual test comparison should be simpler to run and not require an evaluation environment as complex as the one used for the initial test evaluation. Some of this is out of our desire to allow for custom user comparison functions, but given the cost we may want to think of some other way of handling this.

The relevant code is within testItem.