Update Documentation Throughout

[eharpste]

We've made several changes to the core algorithms and so its probably time to update the current documentation. One glaring example is infer_missing claims to return a completed instance when it really returns a tuple where the first element is the completed instance and the second one is the likelihoods of the values in that instance.

[eharpste]

Another big example is the predict module has lackluster documentation compared to all the others.

[eharpste]

I took a pass through the documentation currently on read the docs and here are various issues I noticed

• [x] 1. Update the links to the trestle article on the main page and elsewhere to point to the journal version. The one in the second paragraph of the overview is wrong.
• [x] 2. Update the instance representation page to say attributes must be strings or tuples. I instead updated it to say they must be hashable and zero index-able.
• [x] 3. Figure out how to get the get_best_operation images to actually appear on readthedocs
• [x] 4. check if get_prediction_probability in CobwebNode is still right. It mentions probability based on the root but appears to use past performance, which is a bit different. Same for get_probability()
• [x] 5. Unify how we refer to instances when they appear as parameters, ideally link to the instance representation page.
• [X] 6. document the limit_by_performance parameter of output_json in CobwebNode. We actually just took it out along with all of the past performance stuff.
• [x] 7. document the allow_non parameter of predict in CobwebNode
• [x] 8. typo in infer_missing documentation of CobwebTree (wither instead of either) also the docs for the method seem broken in general.
• [x] 9. figure out why concept_formation.cobweb.random() shows up in the docs. This seems to happen in all the tree classes.
• [x] 10. some kind of latex bug in the docs for expected_correct_guesses() in Cobweb3Node
• [x] 11. Cobweb3Tree claims that scaling is off by default but then says that scaling is set to .5 because it is a rational choice to balance against nominals.
• [x] 12. Not a doc problem but does the DummyTree handle structure mapping correctly under the new format? Updated how DummyTree handles structure mapping, but not sure if the entire class is correct. Probably need some kind of test.
• [X] 13. Actually document the predict module, with all parameters. I actually removed the module and changed it to evaluation to make it clearer what the module was really doing.
• [x] 14. Update references to StructureMapper's "standard pipeline" to reflect the new reality that it only contains the NameStandardizer.
• [x] 15. Either document or hide the StructureMapper's various functions. Many of them either have documentation for params that have been removed or don't document new params that have been added.
• [x] 16. Links to py_search in the structure mapper are formatted wrong, also probably need to be updated to reflect changes in search.
• [x] 17. explain the gensym parameter in the NameStandardizer's constructor.
• [x] 18. The preamble paragraphs to TrestleTree's docs are generally out of date and should be re-written at least in that they refer to a new missing alpha parameter.
• [x] 19. TrestleTree still has a train_from_json() function? No it doesn't but the docs think it does!
• [x] 20. The continuous_value module does not appear in the docs.
• [x] 21. Add module level docs to all the modules.