Better test coverage reports

[antonpirker]

Our coverage reports are broken. This PR tries to fix them.

• Sometimes the coverage report XML files contain references to files in /tmp/... (this can happen if dependencies write those files) so the first change is to omit those files.
• We created our coverage reports with coverage xml -i where the -i means "ignore errors". This is why we never found out about problems generating coverage reports. Report generation fails now verbose (everywhere except in Python 3.6, because there are always some errors there because it can not parse python files with async code, but I guess those can be savely ignored)
• For Python 3.6 we know have a special coverage config (.coveragerc36) because the option exclude_also was named exclude_lines in older coverage.py versions.

[codecov]

❌ 4 Tests Failed:

Tests completed	Failed	Passed	Skipped
11541	4	11537	1637

View the top 3 failed tests by shortest run time

tests.integrations.aws_lambda.test_aws test_init_error[python3.9]

Stack Traces | 10.9s run time

.../integrations/aws_lambda/test_aws.py:334: in test_init_error
    (event,) = envelope_items
E   ValueError: not enough values to unpack (expected 1, got 0)

tests.integrations.aws_lambda.test_aws test_init_error[python3.11]

Stack Traces | 11.3s run time

.../integrations/aws_lambda/test_aws.py:334: in test_init_error
    (event,) = envelope_items
E   ValueError: not enough values to unpack (expected 1, got 0)

tests.integrations.aws_lambda.test_aws test_init_error[python3.10]

Stack Traces | 16.4s run time

.../integrations/aws_lambda/test_aws.py:334: in test_init_error
    (event,) = envelope_items
E   ValueError: not enough values to unpack (expected 1, got 0)

To view individual test run time comparison to the main branch, go to the Test Analytics Dashboard