fix(ds4): honor sampler settings in decode

[dusterbloom]

Summary

• thread GenerateRequest.sampler into DeepSeek4Backend::generate_impl
• sample from logits when temperature or penalty settings are active
• keep the existing greedy argmax fast path when the sampler is inert

Why

DeepSeek4 single-device decode was always using greedy argmax and never reading the request sampler, so temperature / top-k / top-p / repetition settings were silently ignored.

Validation

• git diff --check
• manual source inspection of the decode path

[Graffioh]

hey, here -> #507 in the commits i also tried to fix the sampler + broken greedy decode - can you take a look? let me know if i missed something, trying to learn while contributing

i know is a part of a bigger PR, so since i need a bit of time to test it, i/you can cherry pick that changes and put it in another independent PR