Evaluating cons literal causes panic about lurk-rs

[huitseeker]

isn't this solved by #201 / #224 ? /cc @emmorais @porcuquine

[porcuquine]

Actually, no. There are two ways to fix the manifest problem, both of which need to be done.

The first is that a runtime error shouldn't cause the REPL to exit (with error). This is what #217 was going to fix. I'll make a PR fixing this in just a minute. [UPDATE: #236]

The second is that #201 doesn't actually cause evaluation of (1 . 1) to be an explicit error. It did revamp how those errors are handled, but didn't fix the case named in this issue. (It's very hard to tell from the diff whether it fixed any case, and no test was added. If we identify that some behavior was changed, we'll need to add a test and also ensure it's fixed in the circuit. I'm planning to work on the errors generally, and will include this in the set of case that need fixing in and out of circuit.)