C++ wrapper plans

[dankamongmen]

@grendello did a great job early in Notcurses development with the initial C++ wrappers, and showed me some techniques -- I'm eternally appreciative. He came out of nowhere and dropped a complete set of high-quality wrappers, done pretty much entirely without my feedback, unrequested and unexpected.

If he ever returns to continue maintaining them, I'd be more than delighted to have him. He's been unreachable for a few months, though, in the way that people who don't have the bandwidth to deal with Yet Another Thing become unreachable. In no way am I anything other than thankful. With that said, the C++ wrappers are in need of some love.

Most critical in my mind is eliminating the map<Plane> that attempts to ensure a bijection between ncplanes and ncpp::Planes. i understand what Marek was going for here, but i think it unsound. It leads to problems with destructors and other issues, and certainly complicates the wrappers. all it really seems to buy us (and here's where i'm maybe missing something) is that Planes form a graph isomorphic to the ncplanes--i.e. you can compare two Plane pointers and they'll be the same if they refer to the same ncplane. i don't see that as terribly valuable, and certainly not worth the problems.

so i'd like to rip that out, which ought resolve several open bugs. i'd then like to extend the C++ wrappers out to cover new functionality and match recent deprecations.

open c++ bugs:

• remove Plane map from NCPP #1983
• need a way to transfer ownership when creating widgets #1702
• C++: Plane() constructor is calling ncplane_dup, not ncplane_create #1474
• add ncdirect_core_init() and notcurses_core_init() to c++ wrappers #1312
• C++ get_stdplane() lends itself to aborts following nc.stop() #645

[WSLUser]

This seems a good spot to ask if and how allocations are done with the C++ wrappers. Do the wrappers create any allocations or does it all come from the original C code? The reason I ask is I randomly stumbled upon an MS project that boosts performance of allocations for C++ applications. https://github.com/microsoft/snmalloc It works on Windows, Linux, and BSD, even Dragonfly so seems like it would a perfect useful library to use if it's applicable here.

[dankamongmen]

i wouldn't have expected them to do any allocations, but as it turns out there is a Plane object corresponding to each ncplane see #1983. regardless, it's by no means allocation-heavy. really, nothing is -- all our allocations come in pools.

[WSLUser]

I figured it wouldn't really make sense here but wanted to be sure. I did bring it up for Contour and Windows Terminal, who most certainly do a ton of allocations for various things.