Add Waveshare ESP32-S3-Touch-AMOLED-2.06 board port

[samestrin]

Summary

Adds a board port for the Waveshare ESP32-S3-Touch-AMOLED-2.06 — a 410×502 watch-form-factor AMOLED kit. The port is a hybrid of the existing ports:

• CO5300 display + AXP PKEY for PWR (like the 2.16 port)
• FT3168 touch + fixed orientation (like the 1.8 port)
• No IO expander — LCD_RESET and TP_RESET are direct GPIOs pulsed in board_init()
• 32 MB flash with default_32MB.csv partition table

Pin assignments were taken from Waveshare's reference Mylibrary/pin_config.h in their demo repo.

CO5300 col_offset gotcha (worth highlighting)

The 2.06 panel exposes its visible 410-wide viewport at a horizontal offset inside the CO5300's internal RAM. Without passing the right col_offset1 to the Arduino_CO5300 constructor, framebuffer writes land at the wrong columns and a vertical strip of stale boot content shows through on the right edge. Waveshare's library uses col_offset1 = 22; this port uses 23 after fine-tuning for centering on one unit. The 2.16" port doesn't have this issue because its 480×480 viewport fills the controller's RAM. I've added a note to docs/porting/adding-a-board.md so the next CO5300 porter doesn't lose an afternoon to this.

Rebased onto current main

This branch was originally cut before the sound HAL, the hold-to-pair gesture, and the BLE supervision-timeout fix landed. It's now rebased onto current main, which required three additions to the port beyond the original hardware bring-up:

• sound.cpp — main.cpp now calls sound_hal_init/tick/play_reset unconditionally, so the port wouldn't link without it. The 2.06 does have an ES8311 codec on the shared I2C bus, but the amp path is unverified on this board, so this no-ops (same posture as the C6 1.8 port). The shared chime.cpp engine is ready to wire up once someone tests it on this hardware.
• power.cpp — added power_hal_pwr_long_pressed() / power_hal_pwr_released() for hold-to-pair, mirroring the 2.16's AXP PKEY implementation (LONG + POSITIVE IRQs, plus the 8s press-off bump).
• platformio.ini — picked up the PPCP connection-parameter flags from fix(windows): end the BLE reconnect churn — widen supervision timeout (2s→6s) + crash-proof the daemon #106 so this board doesn't ship with the Windows supervision-timeout churn.

An earlier revision of this PR also tweaked flash-mac.sh to take an optional env argument. #30 fixed that same footgun more strictly on main, so that change is dropped — this PR no longer touches flash-mac.sh.

Test plan

Verified on hardware (pre-rebase build):

• Display renders correctly with col_offset1 = 23 — splash animation centered, no edge artifacts
• AXP2101 PMU init OK, battery percent reads correctly
• QMI8658 IMU init OK (rotation disabled — watch enclosure is fixed orientation)
• FT3168 touch — basic events fire (tap-to-toggle splash) and coordinates are accurate (BLE reset zone hit-area works)
• BLE pairs as "Clawdmeter", daemon (macOS) connects and pushes usage data successfully

Verified post-rebase:

• pio run -e waveshare_amoled_206 builds clean (1.34 MB, 10.2% of the app partition, 31.6% RAM)
• All four existing envs still build (216, 18, 216_c6, 18_c6)

Not yet exercised on hardware:

• Hold-to-pair on the 2.06. The gesture postdates the original bring-up, so the PKEY LONG/POSITIVE IRQ path hasn't been run on this board. The AXP2101 is the same part at the same address as the 2.16 and the button wires straight to PWRON, so I'd expect it to behave identically — but flagging it rather than implying it was tested. Happy to confirm once I have the device in front of me again.

Photo of the working device available on request.

[HermannBjorgvin]

Do you have a picture of it running on the device?

[HermannBjorgvin]

Thanks for the port! I'll merge this once the conflicts are resolved, looks good to me!

[samestrin]