Race Condition in Console Runner

[jugglinmike]

A race condition in the console runner's threading logic causes it to run some
test cases more than once and to not run others at all.

This example demonstrates a test run using 2 parallel threads (via -j 2). I'm
using a shell pipeline to highlight repeated runs:

$ ./tools/packaging/test262.py --command js language/reserved-words -j 2 | sort | uniq -c | sort | tail
2 language/reserved-words/7.6.1-7-10 passed in non-strict mode
2 language/reserved-words/7.6.1-7-11 passed in non-strict mode
2 language/reserved-words/7.6.1-7-1 passed in non-strict mode
2 language/reserved-words/7.6.1-7-8 passed in non-strict mode
2 language/reserved-words/7.6.1-7-8 passed in strict mode
2 language/reserved-words/7.6.1-8-3 passed in strict mode
2 language/reserved-words/7.6.1-8-6 passed in strict mode
2 language/reserved-words/S7.6.1_A1.1 failed in strict mode as expected
2 language/reserved-words/S7.6.1_A2 passed in non-strict mode
3 language/reserved-words/7.6.1-8-1 passed in strict mode

Executing the tests within a single thread (via -j 1) does not present this
problem:

$ ./tools/packaging/test262.py --command js language/reserved-words -j 1 | sort | uniq -c | sort | tail
1 language/reserved-words/7.6.1-8-9 passed in non-strict mode
1 language/reserved-words/7.6.1-8-9 passed in strict mode
1 language/reserved-words/S7.6.1_A1.1 failed in non-strict mode as expected
1 language/reserved-words/S7.6.1_A1.1 failed in strict mode as expected
1 language/reserved-words/S7.6.1_A1.2 failed in non-strict mode as expected
1 language/reserved-words/S7.6.1_A1.2 failed in strict mode as expected
1 language/reserved-words/S7.6.1_A1.3 failed in non-strict mode as expected
1 language/reserved-words/S7.6.1_A1.3 failed in strict mode as expected
1 language/reserved-words/S7.6.1_A2 passed in non-strict mode
1 language/reserved-words/S7.6.1_A2 passed in strict mode

We have a couple options for how to handle this: fix the bug or remove the
optimization. Patches and rationale incoming.