test for possible code generation bug in Petit/Nasm

[WillClinger]

We aren't distributing a Petit/Nasm version at this time, but the Petit/Nasm back end is still part of the code base and might be revived some time in the future.

The next time we build a Petit/Nasm version of Larceny, we should test to see whether it has the bug documented in ticket #688, where the IAssassin back end generated incorrect code for lambda expressions that close over exactly 31 free variables.

That bug has been fixed in the IAssassin back end, and was not present in the other back ends we're using at this time. We don't know whether the bug is present within the Petit/Nasm back end.