fix: resolve first-board bring-up issues (power, OLED, XBee diagnostics, button naming)

[thePunderWoman]

Summary

Everything found and fixed while doing hands-on bring-up of the first assembled board:

• Power-button auto-shutdown: the sense pin's polarity was assumed active-high but is actually active-low (external pull-up to 3V3, switch pulls to GND when held) — firmware read the button as continuously held from boot and triggered graceful shutdown 3 seconds into every session.
• OLED crash when not present: Adafruit_SSD1306::begin() never actually validates I2C ACKs and never allocates its frame buffer unless called — so when the display genuinely isn't there, any later render() call wrote through a null pointer (StoreProhibited panic → boot loop). Added a real I2C presence probe (address 0x3C with 0x3D fallback), a full bus-scan diagnostic when neither responds, and a readiness guard on every OledDisplay method.
• Serial never reached the USB port: this board has no external USB-UART bridge chip, and ARDUINO_USB_CDC_ON_BOOT wasn't set, so Serial was silently going to unconnected UART0 pins instead of the native USB-CDC port used for flashing/monitoring.
• Early boot diagnostics getting lost: native USB-CDC has to fully re-enumerate with the host after every reset, which took longer than the firmware needed to reach its first prints — lost even with a monitor already open and watching. Added a bounded (2s) wait for an actual host connection before continuing.
• XBee bring-up diagnostic: added a boot-time read of the ATTN pin to help separate SPI wiring issues from module/firmware issues in ongoing debugging.
• Left/Right button naming was backwards: kLeftUp/kLeftDown and kRightUp/kRightDown were mapped to the physically wrong button pairs (Vol/Trigger swapped) — confirmed against the real controller and corrected. Nothing else in the codebase hardcodes the physical mapping, so this one change fixes it everywhere (menu-open combo, uplink suppression, Button Test display).
• Button Test screen never updated after the first press: the display's redraw logic only watched currentScreen()/selectedMainMenuItem(), not lastTestedButtonIndex() — pressing a button while already on that screen changed neither, so the screen silently never refreshed.
• KiCad: the OLED Qwiic connector's pin assignment in SnipsControllers:SSD1306_OLED_4pin was wrong on two counts, both found by testing against real hardware — VCC/GND were swapped, and once fixed, SCL/SDA turned out to be swapped too. Both corrected in the symbol (library file + schematic's embedded copy); the PCB itself still needs Update PCB from Schematic (F8) and a small copper rework near the connector for the next revision.

Test plan

• pio test -e native — 262/262 passing
• pio run -e esp32s3 — builds clean
• All fixes verified against real hardware this session: power-button hold no longer triggers spurious shutdown, OLED boot splash renders correctly once the connector was rewired to match the corrected pinout, Serial output reliably visible from the very first boot line, Button Test screen updates live and shows the physically-correct button names
• XBee SPI communication still failing (ATTN pull-up confirmed intact via the new diagnostic) — separate ongoing investigation, not addressed by this PR

🤖 Generated with Claude Code