Improved core testing around gitlab repo interaction

[sb10]

Currently, the gitlab repo address and credentials are taken from a fixed location in vault.

To safely test core, either manually or in automated tests, a developer currently hacks the code to replace vault usage with hard-coded strings, to use a developer's own test repo and credentials.

This is fraught with the possibility of a bad commit that forgets to revert the hack, and isn't an ideal way of working.

Is there a fool-proof way of ensuring that when manually or automatically testing, we use different user-specific values in vault than what production uses?

With that in place, it would then be good to improve the tests, which currently mock repository pushes so they do nothing, and rely on the developer manually creating some test environments in their repo. It would be better to have some test setup and teardown that deletes everything in the repo and then adds files for a desired state ready for testing. But this is obviously dangerous if there's any possibility the production repo could get used.