Non-quantised VQ-VAE LDM training needs modification of the inferer to be able to modify the quantised flag.

[virginiafdez]

A recent PR made it possible for the VQ-VAE encode_stage_2_inputs to take in a flag that allows to operate the LDM with non-quantised encoding. However, this flag should also be passed to the inferer, as that function is called internally inside of the inferer functions.