propose alternative derivation mechanism

in sphere-json & sphere-mongo we propose a generic derivation of typeclass instances based on scala macros & core generation.

I experimented using another derivation mechanism based on [magnolia](https://github.com/propensive/magnolia): https://github.com/sphereio/sphere-scala-libs/tree/derive_with_magnolia

I could use this new mechanism for sphere-mongo in our private backend.

Using magnolia instead of our own macros is a trade-of decisions:
(+) we can mutualize our efforts with other contributors to have a better derivation mechanism
(+) Magnolia advertises having good compilation time
(+) we could only rely on scala annotations and fix https://github.com/sphereio/sphere-scala-libs/issues/131
(+) less code in this repository
(-) depending on magnolia
(-) less liberty

I personally think that this way is worth a try.
But I'd like to avoid a big-bang release where we move completely from our macros to magnolia.

My proposal:
- phase 1: split out the current libraries to extract the current derivation
- phase 2: add magnolia as a possible alternative. The 2 derivations mechanisms coexist. It means that we have to duplicate the tests. But this allow to quickly switch which mechanism we want to use
- phase 3: depending on the results of phase 2, we could keep only magnolia

For the transition, we have 2 possibilities. For example if someone is using `sphere-mongo`:
- 1st possibility: `sphere-mongo` can be a library depending on `sphere-mongo-core` & `sphere-mongo-macro-derivation`.
- This change should be transparent for the user.
- Users have to switch to `sphere-mongo-core` & `sphere-mongo-magnolia` to test in the phase 2.
- 2nd possibility: we extract the derivation from `sphere-mongo` into `sphere-mongo-macro-derivation`.
- We don't need any `sphere-mongo-core`.
- Current users have to add this dependency if they rely on the derivation feature.
- Users are then aware that we will propose `sphere-mongo-magnolia` in the phase 2 as an alternative to `sphere-mongo-macro-derivation`.

Start by reviewing the existing derivation implementations for sphere-json and sphere-mongo, then inspect the derive_with_magnolia branch linked in the issue. Compare the current macro and core-generation structure with the proposed Magnolia approach. Done would require agreeing on the extraction and coexistence design before implementation.

Written by the indexing model from the issue text.