Add callable parameter based interface

[SebastianM-C]

Checklist

• Appropriate tests were added
• Any code changes were done in a way that does not break public API
• All documentation related to code changes were updated
• The new code follows the
  contributor guidelines, in particular the SciML Style Guide and
  COLPRAC.
• Any new documentation only uses public API

Additional context

Implements #54 (comment) and makes the lux model a callable parameter in the NeuralNetworkBlock (as mentioned in #51 (comment))

Needs JuliaSymbolics/Symbolics.jl#1508 (for 774aba9 only. If this is an issue I can split this PR in 2).

@ChrisRackauckas Is the new interface okay?
I might need a bit of help on how to write the parameters more generically, since right now the names are hardcoded.
I suspect I can't use the macro form.

I was also thinking of having get_network work on an ODESystem and retrieve the neural network in the NeuralNetworkBlock case, but I'm not sure how can I check that the ODESystem is a NeuralNetworkBlock. We could just assume it is and just reach out for the expected name and hope for the best 😅

[SebastianM-C]

I fixed the hardcoding of the symbolic variable names and tests pass on the latest release (and fail on pre due to a lot of upstream issues).

[ChrisRackauckas]

This was missing docs. We should document this.

[SebastianM-C]

Ah, yes, I added the docstring but I didn't add it in the actual doc pages 😅

[ChrisRackauckas]

Also a quick tutorial.

[TorkelE]

Looks cool! Once there are some docs I will look at changing my Catalyst implementation to use this instead.