feat: add a randomizer proxy to test (some part of) the testsuite

[laurentsenta]

The worst thing that could happen to us is introducing a bug that disables one or more checks.

An example: in Kubo, due to a typo in the test, we miss a bug ipfs/kubo#9651

It's not crazy to imagine something similar could happen here.
The suite would be green, but the code wouldn't be tested.

We also need extra care when we port kubo tests: we can't do red/green testing.
So as a test writer, I have to introduce bugs (typos to values, wrong payloads, etc) to start from a red test and make sure it's working.

This PR proposes a "randomizer", which lives in front of a gateway and randomizes its outputs (headers and body).

If we run the test suite against our randomizer, we expect every test to fail most of the time.
If the entire test suite passes, there is a problem.

• This is the first step towards "testing the test suite"; it requires little code & maintenance, but we can validate that we catch the most significant mistakes during local dev and in CI.
• We implicitly get some benefits of red/green testing because we know that tests are actually testing things.
• We could introduce more subtle bugs in payloads, content types, etc if we want to

Next Steps

• validate the approach.
• clean the code.
• create an action that runs the test suite against the randomizer gateway and fails if the test suite passes.
• create an action that raises warnings if a test passes.

[laurentsenta]

@lidel @galargh I'd like to get your feedback on that one,
it doesn't replace unit-tests but having a no-maintenance-catch-big-mistakes test would help me "sleep better" :)

[galargh]

I like it! That's really clever ❤️

[galargh]

Maybe we could run tests against random gateway a couple of times. Or rerun the tests that passed. Then, if we found a test case that passed on all the attempts, we'd know something's wrong.