refactor: set /dev/gpiochip0 as default

[2bndy5]

resolves #1017

This also adds some info to the docs that was requested in #1017

[2bndy5]

I'm looking at this more in depth. I think we can get rid of the GPIOChipCache::chipInitialized member. Before this patch, chipInitialized was used to prevent changing the GPIOChipCache::chip string when using more than 1 radio object in the app (ie using multiple calls to pinMode()).

Furthermore, I think we can also get rid of GPIOChipCache::chip and just use RF24_LINUX_GPIO_CHIP directly.

[2bndy5]

I also figured that the cached FD for the GPIO chip could be static (e8269f6). This allows the gpio.cpp and interrupt.cpp modules use the same exact FD for accessing the GPIO chip. 🎉

[2bndy5]

LGTM

This works as expected on my RPi3 and RPi4 (both 64-bit). I specifically tested the interrupt example because it uses all the code changes in this patch. I also ran a modified getting-started example in which 2 radio objects were init'd (second radio obj failed as expected) and a third pinMode(21, OUTPUT) was called from the userspace.

[TMRh20]

Quick tests on RPi5: 64-bit & RPi4: 32-bit are working. Good job on these changes!

[2bndy5]

I probably could've used the static trick before, but the pinMode() wrapper isn't intended to be used publicly; the main reason I think the drivers' implementation should be namespaced (#926).