[Text Gen UX] top level constructor aliases + code gen subclass

[bfineran]

This PR adds the requested constructor aliases for the main text generation tasks. also adds a code generation subclass to reference - idea here is that if there is ever extra processing or defaults requested for code generation, it will be easy to add in to the subclass

requested UX:

from deepsparse import TextGeneration, CodeGeneration, Chat

text_pipeline = TextGeneration(model=MODEL, **kwargs)
text_result = text_pipeline(prompt=PROMPT, **kwargs)

code_pipeline = CodeGeneration(model=MODEL, **kwargs)
code_result = code_pipeline(prompt=PROMPT, **kwargs)

chat_pipeline = Chat(model=MODEL, **kwargs)
chat_result = chat_pipeline(prompt=PROMPT, *args, **kwargs)

test_plan:
manually verified:

from deepsparse import CodeGeneration

code_pipeline = CodeGeneration(model="/home/benjamin/neuralmagic/models/codegen_mono-350m-bigpython_bigquery_thepile-pruned50/deployment", engine_type="onnxruntime")
code_pipeline(prompt="def hello_world:", generation_config=dict(max_new_tokens=10))

[dsikka]

LGTM