Update readme examples to use newer Qwen2 model

[jncraton]

Qwen2 generally outperforms Qwen1.5 according to released benchmarks:

Datasets	Qwen1.5-0.5B-Chat	Qwen2-0.5B-Instruct
MMLU	35.0	37.9
HumanEval	9.1	17.1
GSM8K	11.3	40.1
C-Eval	37.2	45.2
IFEval (Prompt Strict-Acc.)	14.6	20.0

This model is working fine for me, and I would expect it to generally provide a better experience for users exploring this package. I would expect speed to be nearly identical given the similar sizes of these two models.

[abetlen]

Thank you!