Pyproxy mixins

[hoodmane]

This PR improves several things about PyProxy:

1. Reduce stray properties to the absolute minimum. As much as possible attribute access should go to the Python object.
2. Extra methods defined on Proxies of particular types of Python objects should overridden by properties on the python object (as discussed in ENH PyProxy getattr/etc traps & added getitem/etc for PyMappings #1175)
3. Share prototypes between PyProxies of the same type to reduce memory usage.
4. Organization, explanations, more consistent behavior.

Prior to this PR, fields called "name", "length", or "prototype" all showed weird behavior ("prototype" would be shadowed entirely, "name" and "length" show inconsistent results with the "in" keyword). Likewise "caller" and "arguments" were in all cases shadowed. Now the "name" and "length" inconsistencies are fixed, "prototype" is no longer shadowed (and setting and deleting the "prototype" field works as expected) but "prototype" shows some minor (and unremoveable) inconsistency with the "in" keyword. "caller" and "arguments" are now only shadowed if the PyProxy wraps a callable Python object, in other cases they behave normally.

I also added a lot of comments and organized and simplified the logic for testing which abstract object protocols the Python object supports.

This is on top of #1175.

[phorward]

Hello @hoodmane ,
thank for this pull request, it looks very impressive. From my first view I don't see any problems, the tests look nice on the mixins and the documentation is very well. The only "little" thing which bothers me are the comments about when you're not sure about something. I think these kind of thoughts must not be part of the source code, but that's only my opinion.

Separating the Javascript-parts from pyproxy.c into a new pyproxy.js could also be done within this pull request, but IMHO having all related C-JS-stuff in one file is okay.

[phorward]

Addresses #1049, #1006, #693, I'll do a test soon on my original issue 👍

[hoodmane]

@phorward Thanks for the review.

I don't think this PR addresses the GC issues. I'll probably do that after this is merged.

[phorward]

I don't think this PR addresses the GC issues. I'll probably do that after this is merged.

Yes, I found this already out, and I'm very excited about that change.

[phorward]

Thank you @hoodmane that is ok for me. Let's wait for CI, and can you please merge the current master into your branch?

[hoodmane]

So what's the status on this? Is this okay to merge @phorward?