feat(timelines): add transitions

[nrealus]

This PR adds an analysis::transitions module to aries-timelines.

A transition is defined as either a condition, effect, or a pair (condition, effect) where the condition and effect share their source (container action), presence literal, and state variable.
The module notably allows:

• collecting the transitions in a low-level planning model (i.e. SchedEncoder), over "unambiguous" state functions (see nonsimple::collect_nonsimple_conditions_and_effects_to_relax).
• evaluating a transition's terms given a grounding of its source (i.e. an assignment to its container action's parameters)
• collecting the (possible) lifted supports between transitions, which include causal links identified in the planning model (but are not limited to them)

[arbimo]

I am not sure this should be merged as a module independent of the lprelax work.

It seems designed to very tightly accomodate the needs of lprelax, and is probably not useful outside of it.

Independtly of that, the module currently lack high-level documentation: what it does ? why ? and how ?

In my understanding, identifying transitions is just a matter of pairing one condition with an effect that share the same source, presence, state variable and end/start timepoint.
This is already not stated clearly.
I am not sure anything beyond that is useful to have as a generally accessible analysis.

One reason for this is that, it seems to me that the non-simple terms are dropped for at least part of the analysis which is probably incorrect outside lprelax. I am also not convinced that the recovering of the closed world assumption effects is worth the complexity it induces.