[Bug Fix] Update torch import reference in bnb quantization (#1902)

# What does this PR do?

Fixes `Import Error` occurring from mismatch of usage between
torch.nn.Module and nn.Module.

server/text_generation_server/layers/bnb.py

@@ -70,7 +70,7 @@ class Linear8bitLt(torch.nn.Module):
         return out
 
 
-class Linear4bit(nn.Module):
+class Linear4bit(torch.nn.Module):
     def __init__(self, weight, bias, quant_type):
         super().__init__()
         self.weight = Params4bit(