test: consider using custom external nvim process

[tmillr]

There are a couple of (potential) issues when using Plenary alone:

1. Tests are not isolated

• Every test in a spec file runs within the same nvim process/instance requiring manual sanitization between tests

2. Plenary runs tests in headless mode

• This is probably not an issue 90%+ the time, but it seems that Neovim's behavior does differ slightly when headless (particularly when it comes to the issuing of messages/prompts/etc.)

Suggestion: spawn an nvim process/job in each test and run the test in that. Either run the entire it() function within that process, or simply communicate with it over rpc (i.e. call fns and get/set values over rpc as needed, but assertions are still executed in the initial Plenary process).