fix(epd7in5_v2): match Waveshare reference framebuffer polarity

[m5r]

Fixes #256.

Summary

update_frame in Epd7in5 (V2) was sending the user buffer verbatim to DTM2 (0x13), causing every framebuffer to render with bit polarity inverted on the actual panel: Color::White (bit 1) showed as black, Color::Black (bit 0) showed as white. Symmetric test patterns hide this; any image with asymmetric content does not.

Issue #256 documents the same symptom — the reporter suggested buffer_mut().iter_mut().for_each(|b| *b = !*b) as a workaround, which is essentially what this PR does inside the driver.

Root cause

Waveshare's reference C demo EPD_7IN5_V2_Display() sends the user buffer raw to DTM1 (0x10) and bitwise-NOT to DTM2 (0x13). The Python driver display() does the same on the wire, just with the opposite user-facing buffer convention (it pre-inverts in getbuffer per its own e-paper world 0=white, 1=black comment).

The panel's DTM2 register expects bit 0 = white per the UC8179 datasheet §22, KW mode with NEW/OLD, DDX=00: {NEW=0, OLD=0} -> LUTWW. epd-waveshare's Color::White → bit 1 encoding means the buffer must be inverted before reaching 0x13.

Changes

• DisplayInterface::data_inverted — new method, streams ~data via a 256-byte stack chunk (no heap allocation, fits the 48 KB framebuffer comfortably).
• Epd7in5::update_frame — writes DTM1 raw + DTM2 inverted, matching the C demo. Writing both forces a full LUTKW/LUTWK transition for every pixel, producing strong contrast.
• Epd7in5::clear_frame — writes DTM1=0xFF + DTM2=0x00, matching EPD_7IN5_V2_Clear. Previously both were 0x00, which goes through LUTWW (stays white) — works for a freshly-powered panel but doesn't force a clean transition from a prior image.

Verification

Tested on real hardware: ESP32-S3 driving a Waveshare 7.5" e-Paper V2 (G) panel via the Universal Driver HAT Rev 2.3, using esp-hal 1.0 + embedded-hal 1.0. Before the patch, a simple clear(Color::White) + Text "hello world" in Color::Black framebuffer rendered as solid black with white text. After the patch, it renders correctly white-bg-black-text without any application-side color swap.

cargo fmt --check, cargo check --lib, and RUSTDOCFLAGS=-Dwarnings cargo doc --no-deps all pass.

Test plan

• Build clean (cargo check --lib)
• Format clean (cargo fmt --check)
• Docs clean (RUSTDOCFLAGS=-Dwarnings cargo doc --no-deps)
• Real hardware: 7.5" V2 (G) panel renders white background + black text with no app-side inversion
• Real hardware: plain 7.5" V2 (non-(G)) — would appreciate a second confirmation from anyone with that variant on hand