fix leaky EOS VM OC executor code mapping

[spoonincode]

executor's ctor creates a shared mapping to the code cache file it is passed,

leap/libraries/chain/webassembly/runtimes/eos-vm-oc/executor.cpp

Lines 146 to 151 in e9ace45

	struct stat s;
	FC_ASSERT(fstat(cc.fd(), &s) == 0, "executor failed to get code cache size");
	code_mapping = (uint8_t*)mmap(nullptr, s.st_size, PROT_EXEC|PROT_READ, MAP_SHARED, cc.fd(), 0);
	FC_ASSERT(code_mapping != MAP_FAILED, "failed to map code cache in to executor");
	code_mapping_size = s.st_size;
	mapping_is_executable = true;

but its dtor doesn't unmap that mapping. That's sloppy and leaky. Easy enough to fix.

I'm not considering this for stable since realistically nodeos only ever creates one of these per thread for the entire lifetime of the application anyways.