[LibFuzzer] Fix flakey ``fuzzer-trace-pc.test`` test on OSX

[delcypher]

Summary

[LibFuzzer] Fix flakey `fuzzer-trace-pc.test` test on OSX

When running LLVMFuzzer-FullCoverageSetTest-TracePC the memory limit
is sometimes hit which lead to the test failure.

I'm guessing that much more memory is needed on OSX than Linux because
it takes many more iterations for LibFuzzer to hit BINGO on OSX for this
test. I'm not sure if this is to be expected but given that this test
is relying on program counter values it seems reasonable to expect that
the behaviour between platforms could differ.

[kcc]

Something is very much wrong here. Please find what.
The target itself does not allocate any memory.
Of course, libFuzzer does, and there is asan's quarantine so that some memory usage growth is expected
but if it reaches 2Gb something else is bad.

[delcypher]

For public observers. @kcc, @zaks.anna and I have discussed this and I'm just going to disable this test on OSX for now so that we can setup an OSX buildbot. @zaks.anna and/or @kubabrecka will look into the issue later on this year.

lib/Fuzzer/test/fuzzer-trace-pc.test

 CHECK: BINGO
 RUN: not LLVMFuzzer-FourIndependentBranchesTest-TracePC      -seed=1 -runs=1000000 2>&1 | FileCheck %s
-RUN: not LLVMFuzzer-FullCoverageSetTest-TracePC              -seed=1 -runs=10000000 2>&1 | FileCheck %s
+RUN: not LLVMFuzzer-FullCoverageSetTest-TracePC              -seed=1 -runs=10000000 -rss_limit_mb=0 2>&1 | FileCheck %s