This moves all RX packet processing into interrupt context, removing the queueing bottleneck that previous firmwares experienced. Before this change, we could handle at most ~3 packets before they would start NAK'ing due to lack of buffer space until the next main loop iteration pulled additional packets into the fbNew buffer. Now packets can immediately go into fbNew unless there's a pending frame finalization. This vastly improves the efficiency of the USB protocol, and we need far fewer NAKs.

New READMEs for the examples, and update everything for the new layoutful naming convention.

Now the firmware config flags actually select between several different inner-loops. This will be useful in the future if we want to support different configurations with runtime config options, and right now it's useful for side-by-side comparisons with the various config options.

This makes it easier to move around for future optimizations, and it already gains us a few FPS by allowing the compiler to optimize a little better

This actually works fine, we just weren't initializing it to zero, which borked a bunch of assumptions of course.

It's causing USB problems (reading the string descriptors) and it doesn't help now that I've determined that executing from RAM doesn't significantly speed up the updatePixel loop.

This gives one combined image that is convenient to load either with the test jig or with the Teensy Loader

This fixes some bug that I don't have time to investigate right now. Annoying! I suspect an assumption in usb_mem that I'm forgetting about.

For the vendor code to make it through string conversions unharmed, we should make it part of the printable ASCII subset. Fixes compatibility with xusb when reading WCID descriptor.