Expand UI for AmplitudeModel building

[redeboer]

Perhaps this issue is even more general: both expertsystem and tensorwaves (perhaps even pycompwa) need to agree on the syntax on the recipe, so it's natural to put some sort of recipe module into the expert system.

_Originally posted by @redeboer in https://github.com/ComPWA/tensorwaves/pull/60/review_comment/create_

[redeboer]

@spflueger This is related to #40, as the recipe file contains a selection of the particle list that forms the starting point for the expertsystem.

Since the internals of the expertsystem only functions if the input particle info comes in the form of dicts that assume our old XML format (those dicts are used to populate the edges of the graph), I plan to 'enclose' the expertsystem input (namely a particle list) and output (a recipe file) in the form of data structures that can load and dump to those XML-formatted dicts.

A series of PRs will follow, starting with a ParticleCollection containing Particle instances that can be built from both YAML and XML and then dumps its info into a nested dict. Later on, this internal nested dict structure can then slowly be replaced Particle instances.

[redeboer]

Intended structure/workflow: So the idea is that all recipe I/O and the recipe structure is pinned down in the form of classes. Then the expertsystem becomes an actual dependency of tensorwaves and there's no need to define dict to TF object conversions each time the recipe file structure changes.

[redeboer]

Closed by #454