Enhance some networks to set `bias=False` for conv layers

[yiheng-wang-nv]

If a conv layer is directly followed by a batch norm layer, the bias in the conv layer is not need, because the subtract the mean manipulation will cancel out the effect of bias.

Reference:

In MONAI's Convolution class, the default value of bias is True, and the following networks/blocks cannot configure the bias parameter when using Convolution:

• SimpleASPP
• HighResNet
• VNet
• UNet
• BasicUNet
• AutoEncoder

[yiheng-wang-nv]

Therefore, shall we add bias parameters for these networks/blocks?

Or shall we add a judgement as follow in all related palces?

if norm == "batch" and adn_ordering[0] == "N":
conv_params["bias"] = False

conv_block = Convolution(**conv_params)

@rijobro @Nic-Ma @wyli @ericspod