feat(llama.cpp): consolidate options and respect tokenizer template when enabled

[mudler]

Description

Fixes: #7115
Fixes: #6117

This PR aims at two things:

• llama.cpp backend to respect the use chat tokenizer setting which is already part of the YAML config of the model. This instructs LocalAI to piggyback to llama.cpp for templating, leaving inline templates as an options as well, but not strictly needed anymore.
  This allows for instance a YAML config to be only like:

  backend: llama-cpp
  context_size: 8192
  f16: true
  mmap: true
  name: qwen3-0.6b
  parameters:
    model: Qwen3-0.6B.Q4_K_M.gguf

  Which internally would automatically render as:

  backend: llama-cpp
  context_size: 8192
  f16: true
  mmap: true
  name: qwen3-0.6b
  parameters:
    model: Qwen3-0.6B.Q4_K_M.gguf
  
  template:
    # Enable chat templating from llama.cpp
    use_tokenizer_template: true
  function:
    grammar:
    # Disable LocalAI's engine for grammar rendering
      disable: true 

• moves some of the options that were passed by env as options. this allows to configure everything in the model YAML file and avoids generic envs for all loaded models.

  • use_jinja / jinja: Enable Jinja2 template processing
  • context_shift: Enable dynamic context window adjustment
  • cache_ram: Set KV cache RAM limit (in MiB)
  • parallel / n_parallel: Enable parallel request processing with continuous batching
  • grpc_servers / rpc_servers: Configure distributed inference across multiple workers
    Example:

    name: llama-model
    backend: llama
    parameters:
      model: model.gguf
    options:
      - use_jinja:true
      - context_shift:true
      - cache_ram:4096
      - parallel:2
      - grpc_servers:localhost:50051,localhost:50052

[netlify]

✅ Deploy Preview for localai ready!

Name	Link
🔨 Latest commit	234c2ae
🔍 Latest deploy log	https://app.netlify.com/projects/localai/deploys/690e331f93e3aa000820921d
😎 Deploy Preview	https://deploy-preview-7120--localai.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.