Update numpy version to latest

[rahul-tuli]

Updates numpy support to latest

Needs the following pieces from sparsezoo and sparseml

• Upgrade numpy to 1.21.6 sparseml#1623
• Update numpy 1.21.6 sparsezoo#332

Tested locally using the following code and numpy 1.24, including the changes from above two diffs:

from deepsparse import Pipeline

task = "question-answering"
dense_qa_pipeline = Pipeline.create(
    task=task,
    model_path="zoo:nlp/question_answering/distilbert-none/pytorch/huggingface/squad/base-none",
)

question = "DeepSparse is sparsity-aware inference runtime offering GPU-class"
q_context = "What is DeepSparse?"

print(dense_qa_pipeline(question=question, context=q_context))

Also ran all pipeline and transformer tests! They are green.

The problem only arises in question-answering pipeline due to computation of span_extra_info which involved a ragged sequence, supplying a dtype=object solves this; this also seems feasible because earlier numpy versions(<=1.21) automatically converted a ragged sequence into an array of objects, which entails our codebase is already built to handle that case.