feat: add M5Stack PaperS3 simple-config preset (panel-IC 3004)

[AsLY4]

Adds the M5Stack PaperS3 (ESP32-S3, 4.7" 960×540 ED047TC1-class parallel e-paper, 16-gray capable) as a complete device preset, plus its two panel entries.

Firmware support merged in OpenDisplay/Firmware#159 and shipped in release 2.26.5: panel-IC 3004 maps to FastEPD BB_PANEL_M5PAPERS3. This is the toolbox-side counterpart — no new firmware binary and no new manifest, it reuses the existing bin/esp32-s3-N16R8_full.json.

Changes

• simple-config-presets.json — driverBoards: m5-papers3 (index 31); displays: m5-papers3-960x540 (44, mono) and m5-papers3-960x540-4gray (45, color_scheme 6), identical apart from the colour scheme.
• simple-config-id-registry.json — the three matching id → index appends.
• config.yaml — panel_ic_type 3004 (m5papers3_960x540) and minor_version 4 → 5, mirroring the protocol-side @since 1.5 backward-compatible enum addition.

connectorPins uses a new private token 1047, so the board and its two panels only offer each other — same approach as the Inkplate entries (1010 / 1052). The board follows inkplate-5v2 for the parallel-panel shape (all displayPins 0xff, panel pin map lives in the firmware) and the reTerminal entries for powerDefaults.

.github/scripts/validate-simple-config-presets.py passes locally: simple-config OK: 87 presets; id/index registry stable.

Hardware validation

Every value comes from a physical PaperS3 driven through a 24-scenario acceptance run (MONO and GRAY16 rendering over direct and pipe writes, GT911 touch telemetry, buzzer, battery telemetry, deep sleep with timer wake and image retained, plus a full erase → reflash → reconfigure → first image run). Points worth a reviewer's attention:

• The dataBus block is mandatory on this board. Without an explicit I²C block (SCL 42 / SDA 41, 400 kHz) the firmware falls back to Wire.begin() on GPIO 8/9 — which are EPD data lines here — and rendering corrupts into column artefacts. This was the hardest-won finding of the bring-up.
• Touch flags = 0x6 (SWAP_XY | INVERT_Y): the GT911 reports portrait-native axes; this maps them to the 960×540 landscape, corner-verified on hardware.
• i2c_addr_7bit = 0xff for GT911 auto-probe (0x5D then 0x14, board-revision dependent; 0x5D on the validated unit). The schema documents 0/0xFF as auto-detect, though this is the first preset to use it.
• voltage_scaling_factor = 0xa1 (161): validated to −3.2 % against the Li-ion 4.20 V charge plateau (the enclosure exposes no measurement point).
• dataBus.pullups = 0x0 — unlike the reTerminal entries, this board has external I²C pullups; and buzzerConfig.flags = 0x0 because enable_pin is unused here.

manufacturer_id is 0 (diy) with the name "M5Stack PaperS3", so the name-split path advertises manufacturer M5Stack / model PaperS3 — no manufacturer enum change needed.

Two things for you to decide

1. powerDefaults are skipped on mains power. mergeSimpleConfigToFull() applies a board's powerDefaults only when power_mode !== '2' (index.html:3826-3828). The PaperS3's sensible default is mains/always-on (usb), so its battery ADC (GPIO 3, scaling 161, capacity_estimator 1) is dropped in exactly the default configuration — even though the hardware reports battery voltage fine on USB power. The values do apply as soon as a user picks a battery option.

This isn't PaperS3-specific: it affects any board with a battery ADC selected on USB power (reterminal-e1003 has the same hole). I left defaultPower: "usb" with the powerDefaults block intact rather than guess. Happy to either relax that guard so only the sleep-related fields are skipped on mains (one line, here or separately), or add a device-specific powerOptions entry carrying the ADC inline — though that spends a permanent registry index and adds a row to every board's dropdown. Your call.

2. Sequencing with the hosted firmware. bin/firmware-version.json currently pins 2.26.4 (synced 2026-09-02), while PaperS3 support ships in 2.26.5. Until the sync lands, a user flashing from the toolbox would get a firmware without the 3004 mapping and then apply a config it cannot drive. Nothing to edit here — just worth merging this after the firmware sync, or triggering it first.

Happy to reshape any of this.