- Fix device-selection utils for Jax/Haiku.
- Ensure any GPU is left untouched when the device policy specifies
  that the CPU must be used.
- Fix instructions in both HaikuModel and its tests' setup code to
  achieve the former.
- Note: `jax.jit(..., device=<D>)` is used to prevent Haiku from
  using another device than the desired one; however this argument
  is being deprecated, making this solution short-lived. Since the
  supported Jax version is hard-coded however, we keep this patch
  for now, in wait of future framework evolutions.
- Add the use of the optional, custom `--cpu-only` flag in Haiku
  unit tests. This enabled testing the former changes in an env
  that has a detectable GPU, but an unproper driver version (so
  that any undesired use of that GPU would make the tests fail).

Minor gardening around the package

See merge request !44

- First run unit tests (which are the more problematic).
- Then run code quality verifications (pylint / mypy / black).
- Finally run integration tests.

The rationale to inverting the last two items relative to what was
done prior to this update is to keep the longest-running tests for
the end (rather than re-play a long pipeline due to late-detected
formatting issues).

- Make `Vector` a `Generic[T]` to indicate that stored data arrays
  should have coherent types.
- Make `Model` a `Generic[VectorT]` to indicate that a given model
  class is associated with a unique vector class.
- In `Optimizer` and plug-ins ABCs, specify that inputs and outputs
  should be coherent using `Vector[T]` and `Model[Vector[T]]` type-
  hints.
- In concrete plug-in classes, leave the existing code as-is so as
  not to make it harder to read.

Add support for Jax / Haiku

See merge request !32

* Add proper weights and gradients naming.
* Store the model's nested dict of parameters rather than its
  flattened counterpart.
* Refactor some backend code to improve readability of gradients
  computation.
* Add a public getter to access weights' names, opt. restricted
  to trainable ones.
* Revise `set_trainable_weights` to received lists of names rather
  than indices. Keep support for predicates and PyTree inputs.