Separate Input/Output datatype template parameters

[STFleming]

This PR updates the softmax integration to include the ability to specify input and output template datatype parameters separately.

The main changes are:

• Updated to the HLS code to enable the specification of the separate templates for input/output.
• Compiler integration to enable the inference of the output data types and appropriate codegen.
• Test updates to allow the specification of different input/output datatypes.

I have only tested the quantsoftmax node with the following:

tests/fpgadataflow/test_fpgadataflow_softmax.py::test_fpga_dataflow_quantsoftmax[ifm_dim0-INT8-INT8-simd4-hls] PASSED
tests/fpgadataflow/test_fpgadataflow_softmax.py::test_fpga_dataflow_quantsoftmax[ifm_dim1-INT8-INT8-simd4-hls] PASSED

[STFleming]

Added an additional output type test which seems to be working okay.

tests/fpgadataflow/test_fpgadataflow_softmax.py::test_fpga_dataflow_quantsoftmax[ifm_dim1-INT8-INT16-simd4-hls] PASSED