Move transpose based on backend from hls4ml/model/layers.py to backend-specific areas

[jmitrevs]

Prerequisites

• Test that the bug appears on the current version of the master branch. Make sure to include the commit hash of the commit you checked out.
• Check that the issue hasn't already been reported, by checking the currently open issues.

Quick summary

Backed-specific code should generally not exist in hls4ml/model/layers.py, so the backend-specific transposing of weights should not be part of the standard internal representation (IR).

Details

For Conv2DBatchnorm in the standard IR there is a different transpose made based on the backend:

hls4ml/hls4ml/model/layers.py

Lines 560 to 569 in eb70a61

	if self.model.config.is_resource_strategy(self) and self.model.config.backend.name in [
	'Vivado',
	'VivadoAccelerator',
	]:
	self.weights['weight'].data_unquantized = np.transpose(folded_weights, axes=[3, 0, 1, 2])
	self.weights['weight'].data = self.get_attr('weight_quantizer')(self.weights['weight'].data_unquantized)
	else:
	self.weights['weight'].data_unquantized = folded_weights
	self.weights['weight'].data = self.get_attr('weight_quantizer')(folded_weights)

The standard IR should not depend on the backend. This logic needs to be moved to backend-specific code.