Fix build in OpenBSD

[Rangi42]

Fixes #1111, which was filed in December 2022, regarding building RGBDS v0.6.1 on OpenBSD 7.2. We're now up to RGBDS post-0.8.0 and OpenBSD 7.5. We've also switched from C entirely to C++. If this continues to build on the far-most-common CI targets of *buntu Linux, Windows, and macOS, and this fixes it for OpenBSD, then I think we can safely get rid of defining _POSIX_C_SOURCE=200809L.

[Rangi42]

See the discussion in google/flatbuffers#6205 and WebAssembly/wabt#1360 for what to potentially do about this.

[Rangi42]

Okay, so just removing -D_POSIX_C_SOURCE=200809L breaks the build in Cygwin (tested on Windows 7):

I wanted to try detecting the OS (either to opt into the -D on Cygwin and MinGW, or opt out of it on BSD), but BSD make doesn't support that (although it can be done with cmake, using if(MINGW OR CYGWIN)).

Other possibilities:

1. Require GNU make
2. Close Build fails on OpenBSD 7.2 #1111 as wontfix.

[ISSOtm]

A possible fix would be to stop relying on POSIX APIs, in this case, not using fdopen. I don't know how realistic that would be.

(Why do we use it instead of just stdout, anyway?)

[pinobatch]

Might it be a measure to avoid adding $0D bytes before every $0A byte if writing binary output to stdout and it happens to be in text mode on Windows?