docs: fix stale CLI invocations in LLM inference example

[deBrian07]

Proposed changes## Proposed changes

The example commands in the LLM inference guide no longer run as written, tried copying and pasting them and all failed.

Both scripts in mlx-examples/llms/llama now accept flags only:

• llama.py: --model-path (default mlx_model), --prompt, --few-shot,
  --max-tokens/-m, --write-every, --temp, --seed. No positional arguments.
• convert.py: --torch-path, --mlx-path, --model-name, --quantize,
  --q-group-size, --q-bits, --dtype. Writes a directory (default mlx_model),
  not a single .npz file.

docs/src/examples/llama-inference.rst still passes the weights and tokenizer
positionally, and uses --num-tokens, which no longer exists. This updates the
convert.py call, the three llama.py calls, and the surrounding sentence to
match the current CLI. Prompts, sample output, and timings are unchanged.

This is only a doc change, found when following the guide, no corresponding issues.

Checklist

Put an x in the boxes that apply.

• I have read the CONTRIBUTING document
• I have run pre-commit run --all-files to format my code / installed pre-commit prior to committing changes
• I have added tests that prove my fix is effective or that my feature works
• I have updated the necessary documentation (if needed)

[zcbenz]

Thanks!