add enable jax profiler to run_server

[bvrockwell]

• passes profiler arguments to server launch as optional
• uses instructions provided here for how to port forward from remote to local browser

example:

python run_server.py 
--size=8b 
--batch_size=1 
--max_cache_length=128 
--quantize_weights=false  
--quantize_kv_cache=true 
--checkpoint_path=".../models" 
--tokenizer_path=".../models/tokenizer.model" 
--model_name=llama-3
--sharding_config="default_shardings/llama.yaml"
--temperature=0.5
--sampling_algorithm="weighted"
--enable_jax_profiler=True # defaults to port 9999

Test when enabled on v5e-1:

python deps/JetStream/jetstream/tools/requester.py \
--tokenizer ".../models/tokenizer.model" 
Proxy backend support is not added
Sending request to: 0.0.0.0:9000
Prompt: Today is a good day
Response:  to be

Test when disabled (defaults):

python deps/JetStream/jetstream/tools/requester.py \
--tokenizer ".../models/tokenizer.model" 
Proxy backend support is not added
Sending request to: 0.0.0.0:9000
Prompt: Today is a good day
Response:  to be

[FanhaiLu1]

Thanks for adding it!