Clean-up db sub-crate

[LukeMathWalker]

A little bit of polishing/clean-up of our db-related code:

• encapsulate diesel's errors into an opaque error type (anyhow::Error), so that:
  • domain sub-crate does not have to list diesel as a dependency (nice side-effect: compilation speed-up 🚀)
  • domain remains truly unaware of the underlying persistence-related technology choice (ORM-agnostic).
• remove the closure-based mechanism for queries, making it possible to get a pooled connection with a method;
• use the same DatabaseError struct for all DatabaseError error variants in domain;
• misc micro-improvements.

[colinbankier]

This has made me realize that something must have been lost along the way that I missed - the initial point of the Repo and the run method was to "Run the given closure in a way that is safe for blocking IO to the database" - so that it doesn't block the tokio reactor. It used task::spawn_blocking to do this. Did I miss something that this not necessary any more with all the async ecosystem advances?

Even though this may not the a standard way to use diesel, it was something I cared about, however maybe it's confusing in an example app.

Checking async diesel issue made after review.

[LukeMathWalker]

With the latest version of the async-std runtime (not yet released), we should be able to achieve the same result without any special handling: https://async.rs/blog/stop-worrying-about-blocking-the-new-async-std-runtime/
The tracking PR is async-rs/async-std#631

tide is not using it yet, but it will get it eventually, hence I think it makes sense to simplify the code and get the speedup when that lands.

[colinbankier]

Fair enough - I agree. Let's bring that in when available 👍