Allow TOSA tests to not have quant info

Quantized models might have output without quantization parameters
attached to them. The assert for parameters not being None are removed
and handled in order to allow for that case.
numpy transpose is removed in favor of torch.permute to keep the type
of the output after the operation.

Signed-off-by: Per Åstrand <per.astrand@arm.com>
Change-Id: I0e404062154cefa39f18b5706d72d19cac0e6d73