0.9.2-OTAFIX2.3-BP1.5

Since 0.9.2-OTAFIX2.2-BP1.4

• New boards: Heltec T096, Heltec T1, RAK 3401 (#24 — merged from oltaco's upstream, which calls this same board-support work "OTAFIX 2.3" too)
• screen.c refactored to a display-size-agnostic layout (macro-based bar/text positions instead of hardcoded pixel offsets) — no behavior change for existing boards (#24)
• ST7735 display controller support, alongside the existing ST7789 path (#24)
• Five upstream Adafruit fixes backported, hardware-validated on a RAK4631 (#26):
  • Boot loop from a wrong REGOUT0 read
  • GCC 15 build errors (Makefile version filter, ghostfat.c)
  • -Werror=array-bounds false positive in bootloader_settings.c, now a scoped pragma instead of the previous repo-wide --param=min-pagesize=0 workaround
  • Clear CONTROL/PRIMASK/BASEPRI/FAULTMASK before jumping to the application — was rarely causing lockups
  • Wait for BLE notification-queue room instead of silently dropping DFU completion notifications
• Fixed a latent out-of-bounds write in screen.c's print() (#24 review fix) — not reachable on any current board, but a real hazard for future board.h layouts
• Docs: bank_0-zeroed BLE-OTA-wait behavior on serial DFU, refreshed board/check counts (#27, #28)

Hardware-testing note: Heltec T096, Heltec T1, and RAK 3401 are new to this fork and have not been hardware-validated here — per CONTRIBUTING.md's own bar, "a bootloader that compiles but was never flashed is not a contribution, it's a liability." None of the three are flagged in Meshtastic-Android's device_bootloader_ota_quirks.json as requiring a bootloader upgrade, so this is parity/completeness work, not something blocking anyone. Everything else in this release (RAK4631, and the five backported fixes) was tested on real RAK4631 hardware, including a full BLE OTA DFU round-trip.