IDX-65: Fix the execution of the Cedar integration tests

[patjakdev]

There are myriad different issues addressed in this PR, but the result is that the corpus tests now run successfully and pass:

1. We now use shell: bash in the Github action. This sets the -o pipefail behavior that you might expect to be a default (see https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#exit-codes-and-error-action-preference). Without this setting, go test -count=1 -v -tags corpus ./integration_tests/... | tail will always return a success error code.
2. As written, the Github action would create an issue any time the integration tests failed, even if they were run manually. We really only want to create issues if the automated nightly tests start failing.
3. The number of corpus tests has dropped below our threshold for assuming we've somehow screwed up the extraction of the tests from the published tarfile. I updated the threshold to 100, an order of magnitude below the current count of tests. The commit that reduced the test count appears to have been this one.
4. Fixes a casing error in the comparison with the decision string from the test.

[philhassey]

For lack of knowing the exact syntax of the YML, it looks good. I recommend perusing this PR from the AWS Cedar team, as it includes some other minor tweak relating to this.

cedar-policy#16

[patjakdev]

For lack of knowing the exact syntax of the YML, it looks good. I recommend perusing this PR from the AWS Cedar team, as it includes some other minor tweak relating to this.

cedar-policy#16

@philhassey Thanks. I've incorporated the other changes they made into my PR and confirmed that the tests continue to pass.