💚 Use a unique registry for every test

[randycoulman]

Describe the purpose of your pull request

In order to fix the frequently-failing tests, this takes a different approach in addition to running many of the tests synchronously.

It seems as though our ConfigCat instances aren't fully shutting down and removing themselves from the Registry before the next test starts, even though they should be doing that.

Instead, we start a uniquely-named Registry for each test, injecting it into the code using the ProcessTree library.

We allow tests to start their own Registry and pass it into the start_config_cat helper function for the one that needs to refer to the exact same registry.

To make this cleaner, we extract a via_tuple helper function into a new ConfigCat.Registry module so that we can localize the lookup of the registry to use.

This is an alternative to #142.

Related issues (only if applicable)

Provide links to issues relating to this pull request

Requirement checklist (only if applicable)

• I have covered the applied changes with automated tests.
• I have executed the full automated test set against my changes.
• I have validated my changes against all supported platform versions.
• I have read and accepted the contribution agreement.

[kp-cat]

Hey @randycoulman,
Thanks for the PR. Finally, we accept the other PR (--max-cases 1) because we would like to avoid introducing extra dependencies.