Caja interferes with stack management/events during big-bang

[jpolitz]

This is really hard to debug, so just making a note so this doesn't get lost.

I tried re-enabling Caja for evaluating compiled (and handwritten) source modules. When I do, most programs run fine, but when big-bang runs, the seems to restart immediately, and expect the world value to be an entire module result. The causes problems at the REPL, which tries to access the namespace field and go onto the next REPL entry, when it should be waiting.

I'm not sure if it's a generativity problem (e.g. multiple Cont, PyretException, PauseException instantiations) or that Caja has try/catch frames in the way that interfere; this is what's tricky to debug.

Anyway, it's worth a shot to figure it out at some point, though not critical.