This is getting... somewhere. I couldn't help but come back for another crack at this.

The TLDR is- our build setup wasn't exactly robust and up to date, and this should - with any luck - give us a more rational set of steps that aren't against the grain and hard to debug.

Mostly-

• Stop copying board dirs into the MicroPython and Pico SDK source trees 😬
• Use -DMICROPY_BOARD_DIR instead
• Borrow a trick to make the PIMORONI_TUFTY2040 board let Pico SDK correctly locate pimoroni_tufty2040.h
• Explicitly denote which boards need hacks and local board dirs in the workflow matrix- this shows up in CI runs, so we can tell at a glance with flavour of build it is
• rename everything so it's not so darn wordy, no more MicroPython / Build picow_cosmic_unicorn, it's now MicroPython / cosmic_unicorn
• combine Pico W and Pico style builds into a single workflow so they can share the cached MicroPython fetch

Given my recent tweaks to C++ code to avoid init RAM allocation (and on-init RAM initialisation I guess) I need to re-benchmark the Pico W startup time for Enviro to see if it's still necessary to include a full wakeup GPIO patch, or if we can move to the leaner, overclock-only version we use for Badger 2040.

If it's the latter, the overclock-only patch can just be applied to all builds for the very small, early speedup that it gives prior to proper clock configuration. This hack/patch was originally contrived because MicroPython's gc_heap variable was being zero-init, which took a surprisingly long time (adding milliseconds to the boot). I fixed this with micropython/micropython#9017 but it's subsequently been rendered irrelevant by MicroPython's greedy heap change.

Additionally I learned that C++ init_priority(101) is per compilation unit, so the wakeup GPIO module can be - at least taking into account the quirks of our current compiler, and with the understanding the init order is not guaranteed - moved to earlier in the .cmake file so that it runs earlier on boot. Combined these might be enough to move wakeup GPIO latching from ASM to C++ as per - https://github.com/pimoroni/badger2040/blob/62deec7434a01da330a8a39eb999deb8b013ddb2/firmware/modules/wakeup/wakeup.cpp#L15-L17