Refactor integration tests

[dzelemba]

Filling this as a future cleanup.

Integration tests are currently using a combination of bash scripts & cloud build settings to run a host of tests all part of one global config. This causes a few issues:

• it's hard to isolate individual test cases to understand what is being covered
• changing one test case can break other independent tests (e.g. having to update the build step names, or inserting into the one shared dicomstore)
• cleanup is easy to miss
• diff checks are just based on file comparisons which makes them hard to debug

Instead we can move towards using a proper language to write the tests (like Java or Python) that provides connivence methods for creating/cleaning up test environments. A design doc outlining this would be a good next step.