Neewer-Bridge v1.1.2

Fixed

• One light silently ignored while the others work. If a light's DMX
  channels ran past the end of the data a console was actually sending (a wrong
  address, or a console configured for a smaller universe), the bridge skipped
  that light with nothing in the log — it just sat on its last colour, looking
  like dead hardware. It now warns once naming the light and its channel range,
  and logs again when the data covers it (the light still holds its last state,
  as before).
• Firmware OTA header with a long device name. The 0x96 header's
  single-byte length field wrapped for names of 245 characters or more (the name
  defaults to the firmware filename's stem), producing a frame whose declared
  length disagreed with its contents — which these two-chip fixtures re-frame
  by. The cosmetic name is now trimmed to fit.
• artnet-send with an out-of-range --address. Channel values placed past
  channel 512 were dropped when the packet was encoded, and the command still
  reported a successful send. It now refuses the patch with a clear message.
• test --set with an out-of-range value. Numbers wider than the field they
  land in wrapped silently (…:300 became 44), so the probe reported a value
  it never sent. Out-of-range arguments are now rejected; --set raw:<hex>
  remains the escape hatch for deliberately out-of-spec frames.

Changed

• run validates its configuration up front, so a malformed config reported by
  a library caller is a plain startup error rather than a panic in a background
  task. (Nothing changes for the CLI, which already validated on load.)