Add variable bit width support to ONNXToDeepsparse

[KSGulin]

This PR adds support for variable-bit weight quantization in the ONNXToDeepsparse exporter. This affects two steps:

• Conversion of intiailziers to unit8
• Clipping in quantization of weight arrays

Test Plan Local sparsify runs with int-4 weight quantization

[KSGulin]

@bfineran good callout. Updated the array quantization routine and propagated the bit_width args to address this