Check for existing breakpoints when asserting absent sys.gettrace

[tkcranny]

While developing a new strategy for hypothesis, I've run into the case where the code can't be debugged, at least by simply running pytest in a debugger. This is due to an assert in the conftest file:

hypothesis/hypothesis-python/tests/conftest.py

Lines 80 to 83 in 026fa35

	if not IN_COVERAGE_TESTS:
	@pytest.fixture(scope=u'function', autouse=True)
	def validate_lack_of_trace_function():
	assert sys.gettrace() is None

Personally I'm not sure on the meanings behind its existence and operation, but via a chat with @Zac-HD, it came up that if the (coverage) tests have a breakpoint, then they should allow the tests to proceed.

[Zac-HD]

The idea here is that we are testing that there is no trace function unless we are running everything under coverage in CI, to check that the internal use of coverage metrics isn't leaking.

As a developer experience thing, we should avoid raising AssertionError if an explicit trace function has been set by e.g. pdb.

[Zac-HD]

# This condition is to allow contributors to run our test suite under a
# debugger, without giving up the check for a leaky coverage system in CI.
IN_LOCAL_DEBUGGER = all([
    sys.gettrace() is not None,  # probably needs to specify that it's not coverage either
    os.environ.get('CI') != 'true',  # We're never using a debugger in CI!
])

if not (IN_COVERAGE_TESTS or IN_LOCAL_DEBUGGER):