Remove hardcoded number of output groups and rank

[LifeDJIK]

What changed:
Replaced hardcoded values for n_groups and rank with DS4_N_OUT_GROUP and DS4_N_LORA_O.

Why:
This allows to run Pro models on CPU. With hardcoded values ds4 will stop with an error: "grouped Q8_0 tensor has an unexpected layout".

$ ./ds4 -m DeepSeek-V4-Pro-IQ2XXS-w2Q2K-AProjQ8-SExpQ8-OutQ8-Instruct-imatrix.gguf
ds4: context buffers 631.09 MiB (ctx=32768, backend=cpu, prefill_chunk=0, raw_kv_rows=128, compressed_kv_rows=8194)
Commands:
  /help          Show this help.
  /think         Use normal thinking mode.
  /think-max     Use Think Max only when context is at least 393216 tokens.
  /nothink       Disable thinking mode.
  /ctx N         Set context size for following prompts.
  /power N       Set GPU duty cycle percentage, 1..100.
  /read FILE     Read a prompt from FILE and run it.
  /quit, /exit   Leave the prompt.
  Ctrl+C         Stop generation and return to the prompt.
ds4> /nothink
Thinking mode: none.
ds4> Hi!
processing 11 input tokens: 11/11 (100.0%)
Hello! How can I help you today?
ds4: prefill: 0.62 t/s, generation: 0.57 t/s