Supplement fuzz testing with random memory failures

[squeek502]

One thing that the current fuzz testing doesn't test for is proper handling of out-of-memory errors when allocating/reallocating memory. Could do something like https://stackoverflow.com/questions/21878083/how-to-create-random-memory-allocation-failure to do some fuzz testing with random memory failures to make sure we're properly handling those cases.

[squeek502]

Did some basic testing using FAINT: https://github.com/misc0110/faint

d2itemreaders test cases get through without crashing, so that's good. Might be worth adding to CI in some way.