add api to extract inference trace

[noskill]

Sometimes it is useful to have graph or tree for inference.
Like source -> rule -> rule -> rule -> result

Usecases:

• theorem proving
• planning

[ngeiswei]

The backward chainer already somewhat supports that, i.e. the user can pass an auxiliary atomspace where the trace is stored. This is for instance used in the inference control meta-learning experiment https://github.com/opencog/opencog/tree/master/examples/pln/inference-control-learning

For more info on how to do that see (help cog-bc).

Regarding the forward chainer, the C++ code does some of that, but they are no scheme bindings for it. It could be added, although keep in mind that I eventually intend to merge the forward chainer into the backward chainer code (but that may far in the future, although I hope not).

[noskill]

I found that forward chainer also writes some trace into current atomspace, though it is not possible to find how all premises became satisfied for current rule application.

[ngeiswei]

No inference control meta-learning experiments have been attempted with the forward chainer yet, so trace support is lacking.

[ngeiswei]

The following constitute a start of an API to extract FC traces

https://github.com/ngeiswei/reasoning-bio-as-xp/blob/6006874a9918b29510080d5ef43ffc5b7fc809e4/bio-as-utils.scm#L348-L479

It just needs some work and being moved to the URE.