fix: enable Kraken Elite v2 rendering

[Cesarsk]

Summary

• add the transport and protocol fixes needed for Kraken Elite v2 / PID 0x3012 to actually render bucketed content
• prefer PyUSB for the bulk payload interface while keeping WinUsbPy as a fallback when PyUSB is not available
• rewrite the GIF demo flow so it uses the corrected bucket write sequence and tolerates the device's missing final ack on larger transfers

What Changed

• driver.py
  • add PyUSB bulk-device discovery matched by serial number, with WinUsbPy fallback
  • chunk bulk writes instead of sending one large blind write
  • run the LCD initialization sequence before writes
  • accept the create bucket response quirk where status byte 4 still means success on this device
  • use the corrected write commands for GIF/RGBA transfers (0x36 0x01 <bucket> and 0x36 0x02)
• writeGif.py
  • replace the stale script with a working bucketed GIF write path
  • resize/quantize frames before upload and fit the GIF to the device bucket size
  • clear previous transfers/buckets before writing and switch the LCD back to the written bucket after upload
  • tolerate the missing final completion ack observed on larger GIF uploads

Why

The original code detected the device but did not successfully render on Kraken Elite v2 / PID 0x3012.
On validated hardware, the main blocker was the bulk payload path: the old WinUsbPy-only flow was not reliably reaching the LCD payload endpoint, while PyUSB did.
This change keeps the existing architecture intact and only updates the parts needed to make bucketed GIF/static rendering work on the newer Elite model.

Validation

• validated on NZXT Kraken Elite RGB 2024 / Kraken Elite v2 (VID 1E71, PID 3012)
• confirmed the device accepts control traffic and renders bucketed content after the transport/protocol fixes
• verified python -m py_compile driver.py writeGif.py

Notes

• this PR is intentionally narrow and only covers the code needed to fix rendering support for the newer Elite model
• live long-running fast Q565 streaming was not part of this validation; the confirmed working path here is bucketed rendering

[Cesarsk]

Hi @brokenmass I would like to contribute to this repo, as I was worked on a similar project. Hope it's appreciated; if not, feel free to close it.